Does density drive development?
REHEMA MSULWA
IVAN TUROK
3
Report Series produced by the South African
Research Chair in Development Planning and
Modelling, School of Architecture and Planning,
University of the Witwatersrand.
SUMMARY
There is growing interest among governments and researchers around the world in the contributio­n
of cities to economic development. Several influential international organisations have argued
that the spatial concentration of economic activity is necessary for faster economic growth. This
paper examines whether the density of population and economic activity influences the rate of
local economic growth in South Africa. Municipalities are the basic units of analysis and the time
frame is 1996-2010. Contrary to expectations, no statistically significant relationship is found
between density and growth across the full range of 237 local municipalities. However, searching hard for a relationship among particular kinds of municipality, some evidence does emerge.
The influence of human skills on local growth is also examined and is found to be more robust
than density. Several reasons are given for why the relationship between density and growth is
generall­y weak or non-existent.
ACKNOWLEDGEMENTS: We are grateful to the NRF and to Prof. Philip Harrison, the South African Research
Chair in Development Planning and Modelling in the School of Architecture and Planning at the University
of the Witwatersrand for their financial and other support for this study. This study was prepared as a
contributio­n to the Urban Transformation Research Project (UTRP).
CONTENTS
3
DOES DENSITY DRIVE DEVELOPMENT? REHEMA MSULWA IVAN TUROK
SECTION 1
INTRODUCTION
2
SECTION 2
LITERATURE REVIEW
5
IN JOHANNESBURG
SECTION 3
ESTIMATION APPROACH
12
SECTION 4
EMPIRICAL ANALYSIS: DESCRIPTIVE
19
SECTION 5
EMPIRICAL ANALYSIS: SUMMARY STATISTICS
25
SECTION 6
EMPIRICAL ANALYSIS: REGRESSION RESULTS
28
SECTION 7
CONCLUSION
39
BIBLIOGRAPHY AUGUST 2012
41
APPENDICES 43
1
SECTION
1
INTRODUCTION
There is increasing awareness that the spatial distribution of population and industry is important for
economi­c development. Geography matters in all sorts of ways, including the degree of proximity between
economic agents (‘density’). A growing body of evidence suggests that larger and higher density settlements
experience higher levels of productivity and growth. The World Bank’s 2009 World Development Report
raised the profile of economic geography on the policy agenda of many countries and international organisations. It argued strongly that the steady concentration of economic activity through urbanisation is both
inevitable and desirable for national economic growth. Governments should not try to counteract spatial
inequalities by restricting urbanisation or encouraging development in marginal and dispersed locations.
No country has grown to middle income without industrialising and urbanising. None has grown to
high income without vibrant cities. The rush to cities in developing countries seems chaotic, but it is
necessar­y … urbanisation, done right, can help development more in Africa than elsewhere (World Bank,
2008: 24, 285).
The main objective of the present paper is to assess whether the proposition that density drives development
is applicable to South Africa. Although many international studies have explored this question, to our knowledge similar research has not been conducted in South Africa. We seek to fill the gap by analysin­g whether
density has any influence over the rate of economic development across the country. This is importan­t
for a country seeking to tackle poverty and inequality by maximising employment growth and generatin­g
additional tax revenues from economic growth to fund the extension of essential services throughout the
population.
The concept of external scale economies and the related concept of increasing returns to scale underpin
the argument that density supports development. Density raises productivity and growth because firms
benefit from ‘positive externalities’ in large concentrations of economic activity. These include (i) sharing
infrastructure and information; (ii) matching production requirements such as skills and premises, and (iii)
learning about new techniques, products and services through ‘knowledge spillovers’. Isolated firms may
gain internal scale economies by increasing their scale of production, but they cannot exploit the benefits of
external scale economies.
External scale economies arising from concentrated activity are generally defined as ‘agglomeration
economie­s’. The greater the concentration of firms, workers, suppliers and consumers in one place, the
greater the scope for these interactions to generate advantages for each agent. Different places facilitate
agglomeration economies to different degrees and in different ways. This creates a hierarchy of density,
2
reflecting the uneven distribution of activity at different scales. At the top of the hierarchy is the primary city,
with the greatest economies of scale. At the bottom are small towns and rural areas, with a continuum of
settlements of varying size and density in between.
Other factors besides size and density can also affect the relative trajectories of different cities and towns.
The historical industrial composition of a city may favour the persistence of industry concentration. Places
get ‘locked-into’ particular economic structures and continue to attract disproportionate investment in
those sectors because of the head-start they enjoy through distinct skill-sets and supporting institutions
with accumulated knowledge and expertise. Alternatively, cities and towns may be disadvantaged by some
inherite­d feature, such as an isolated physical location or the depletion of mineral resources on which the
city’s initial growth depended.
This paper assesses the relevance of these ideas in the case of South Africa. This is an interesting and important case, bearing in mind the history of extreme policies of spatial control of population movements and
business location. The paper examines (i) whether and to what extent settlement density has affected the
rate of growth of different places post-Apartheid, and (ii) the nature of the settlement hierarchy across the
country. Local municipalities constitute the basic spatial units of analysis.
To estimate the strength of the association between density and growth, we run a series of regressions and
present these alongside their corresponding scatterplots. Density in 1996 is the independent variable, and
growth over the period 1996 to 2010 is the dependent variable. Regression analysis allows us to estimate the
sensitivity (or ‘elasticity’) of growth in relation to density.
Three measures of density are used: (i) the size of the resident population relative to the physical area of
the municipality (‘population density’), (ii) the level of employment relative to the same area (‘employment
density’) and (iii) the scale of economic output or economic activity (gross value added - GVA) relative to
the same area or (‘economic density’). For the sake of comparison, we also assess the relative importance
of skills (or ‘human capital’) on growth on the grounds that this is another potentially significant driver of
development, independent of density. We measure skills as the proportion of people in each municipality
with matric.
Three measures of growth are used: (i) economic output (GVA), (ii) total employment and (iii) total population. Population can be treated as an indicator of growth at the local level because of its influence on the
consumption of goods and services, including public and private services. Population in the form of labour
is also an input into economic activity. Consequently, expanding populations are generally associated with
growing economies, especially at the local and regional scales.
The analysis also explores variations among different types of municipality (urban versus rural, metros
versu­s secondary cities, former Bantustans versus commercial farming areas, and predominantly manufacturing versus mining and agricultural economies). The strength of the association between density and
growth might be expected to vary between these categories because of their distinctive economic and
physica­l characteristics. For example, unusually high population densities and poor performing economies
might be expected in the former Bantustans compared with commercial farming areas because of their historical experience of forced relocations and restrictions on mobility. Figure 1 (see p6) shows the number of
municipalities in each category.
By analysing municipalities in these different categories, we are further able to infer (i) whether physical
characteristics, historical experience or industrial structure influence the relationship between density and
growth, and (ii) whether the association between skills and growth varies with the industrial composition of
the local economy.
SECTION 1
3
FIGURE 1: The number of municipalities by municipality type
The paper proceeds as follows. Section 2 reviews the relevant literature and evidence. Section 3 describes
the methods and data sources used. Section 4 describes the distribution of density across the country. In
Section 5 and 6 we present the original results of the analysis. Section 7 draws the findings together and
offers tentative explanations.
4
SECTION 1
SECTION
2
LITERATURE REVIEW
2.1. Economic Geography and the World Development Report
The World Development Report (World Bank, 2008) argues that economic progress depends on spatial transformation involving: (i) higher densities as cities grow; (ii) shorter distances as workers and businesse­s
migrate closer to density and reduce the costs of transporting inputs and outputs, and (iii) fewer divisions
as nations lower their economic borders and enter world markets to take advantage of scale and trade in
specialised products (the 3Ds). Density is the most important dimension at the local level. Densely populated places enjoy larger agglomeration economies as firms locate close to each other. These advantages
are offset by other forces that promote dispersal of firms, especially firms that are more sensitive to higher
production costs and congestion in cities (Krugman, 1998; World Bank, 2008). Table 1 summarises the forces
for concentration and dispersal.
TABLE 1: Forces for and against concentration
Centripetal forces (promote concentration)
Centrifugal forces (promote dispersal)
Market size effects: Sites with good access to large
markets are preferred for the production of goods subject
to economies of scale (‘backward linkages’). A large local
market supports the production of intermediate goods,
lowering costs for downstream producers (‘forward
linkages’).
Immobile factors: The location of land, natural resources
and sometimes people may impede the concentration of
production. Producers may have to go to where the workers
are. Also dispersed workers create a dispersed market of
consumers, encouraging some producers to locate close by.
Thick employment markets: An industrial concentration
supports a thick local labour market, especially for
specialised skills, so workers find it easier to find employers
and vice versa.
Land rents: Concentrations of economic activity increase the
demand for local land, driving up local land rents and thereby
discouraging further concentration.
Knowledge spillovers: Local concentrations of economic
activity create external economies via information spillovers.
Pure external diseconomies: Concentrations of activity can
generate external diseconomies such as congestion and
over-heating.
Source: Krugman (1998: 8)
Agglomeration economies can be subdivided into localisation and urbanisation economies. Localisation
economies arise from the interactions between firms within particular industries, including shared inputs
and information, experimentation and mutual learning about similar technologies, and access to common
skill-sets. The ‘clustering’ of upstream and downstream firms and associated institutions such as universitie­s
and trade associations can foster creativity, innovation, competitiveness and accelerated growth.
5
Urbanisation economies arise from business interactions in different industries. Industrial diversity allows
firms to share indivisible facilities or public goods (such as transport infrastructure, logistics systems or
broadband networks), a wider variety of suppliers of equipment, components and services, and a larger
pool of specialised workers, professionals and managers. These positive externalities lower costs, promote
flexibilit­y and facilitate growth. There may also be some cross-fertilisation of ideas between industries,
such that industry-specific knowledge spills over into other industries to become city-wide innovation. This
increases the adaptability of local economies, promotes diversification, and avoids lock-in to outmoded
activities and stagnant markets (Overman and Venables, 2005; Roberts and Goh, 2010; Glaeser, 2011).
Agglomeration economies can be amplified by density. The desire for proximity between firms comes from
the need to reduce transportation costs and other obstacles to the exchange of goods, services, people
and ideas. ‘Network effects’ are another advantage - the more firms in the network, the more information,
knowledg­e and intelligence available to learn from. Dense cities are conglomerations of consumers and
producers, buyers and sellers, firms and workers (Glaeser and Kahn, 2003).
People choose to live close to one another, paying high rents and tolerating crime and congestion. Firms
are drawn to dense areas concentrated with people and infrastructure by the possibility of serving a
large local market from a large plant at low transport costs. Increasing returns-to-scale production technology leads to large factories with many workers. The sizable workforce forms a large local market. By
reducing transport costs, cities with a large local market attract firms in different industries. So a selfreinforcing process of agglomeration that begins with the expanding local market further raises industry
productivit­y (World Bank, 2008: 134).
The World Development Report argues that the spatial concentration of activity is necessary for faster
econom­ic growth. Therefore, low and middle income countries should not divert major resources to try and
narrow spatial inequalities as this will jeopardise economic progress. The report argues that inclusive development is still possible. The key is integration. Three sets of instruments are proposed to address the 3Ds of
density, distance and division. Institutions are universal services governments should provide regardless of
place, including basic amenities such as the administration of justice, health and education. Infrastructure
refers to spatially connective investments, such as railways, roads and telecommunications. Interventions
are spatially focused incentives and investments that favour particular places, such as export processing
zones and slum upgrading programmes. These instruments vary in importance at different points in the
national trajectory of development. Figure 2 summarises how these instruments should be matched at different stages. Above all, governments should promote agglomeration economies by building density, while
reducing the time and costs that threaten to undermine rising concentration through congestion and other
inefficiencies associated with large cities (Munoz et al, 2009).
3D – Build Density, reduce Distance, eliminate Division
• Blindinstitution building
• Connectiveinfrastructure provision
• Targetedslum upgrading/clearnace/relocation.
place specific initiatives
2D – Build Density, reduce Distance
• Blindinstitution building
• Connectiveinfrastructure provision
1D – Build Density
• Blindinstitution building
FIGURE 2. Policy priorities at different levels of urbanisation
6
SECTION 2
Degree
of urbanisation
(type of area)
Advanced
≈ 75%
Intermediate
≈ 50%
Incipient
≈ 25%
Source: (Munoz et al., 2009)
2.2
Overview of the literature
It is difficult to measure the effects of agglomeration because of their complexity and feedback effects, and
because they may be outweighed by other factors. Consequently, there is no consensus about the best
way of doing so. Different studies employ different analytical frameworks, economic variables, estimation
method­s, types of data and spatial units. Such studies can be classified according to three dimensions
of agglomeratio­n economies: (i) their industrial scope, (ii) geographical scope, and (iii) temporal scope
(Rosenthal and Strange, 2004). Across all three dimensions, agglomeration economies are said to exist
when places of higher density are more productive and prosperous (Glaeser and Gottlieb, 2009).
2.2.1. Industrial scope
Studies concerned with the industrial scope of agglomeration economies generally make the distinction
between localisation and urbanisation economies, although this distinction is not always as clear in practice
as it is conceptually. Localisation tends to mean industrial specialisation, which can be measured by the
share of a city’s employment in a particular industry. Competition is also important in localisation economie­s,
measured by the number of firms in that industry. Urbanisation economies should be reflected in industrial
diversity, rather than specialisation. The theoretical foundations can be traced back to Jacobs (1984), who
argued that diversity is a source of externalities as industries can borrow ideas from each other.
2.2.2.Geographical Scope
The standard approach to capturing geography is to use administrative boundaries such as municipalities to define the extent of the city. However, this ignores the distribution of economic activity across the
municipality – whether it is concentrated or spread out can presumably make a difference. It also tends to
assume that all firms in a city benefit from all other firms in the city, regardless of the distance between them.
Finally it ignores the effects of firms immediately beyond the boundary, which is a weakness if municipal
boundaries are tightly drawn. Several studies use smaller geographical units to capture the graduated
effects of agglomeration (Ciccon­e and Hall, 1996) or assess the cross-boundary impacts (Rosenthal and
Strange, 2003; Soest et al, 2002).
2.2.3.Temporal scope
The issue here is whether the analysis is able to detect whether agglomeration economies are static (onceoff ) or dynamic (cumulative), i.e. their temporal scope. For example, Henderson (1997) suggests that learnin­g
from neighbours takes time to develop. Dynamic studies take into account the potential for agglomeration
advantages to accumulate and affect the growth of a city over subsequent periods (Rosenthal and Strange,
2004). They regress growth against agglomeration indicators at the beginning of the period.
2.2.4.Growth variables
Different measures of growth have been used in empirical studies, including increases in productivity, income,
employment, population and output. The choice depends on the specific proposition being examined and
the available data. It may be sensible to use a range of measures to investigate the scope of agglomeration
economies and avoid false conclusions. Of course, careful interpretation of the findings is important since
different variables are not interchangeable (Cingano and Schivardi, 2003; Fafchamps, 2004; Rosenthal and
Strange, 2004). Some studies have used employment growth as a proxy for productivity because of the lack
of suitable firm-level data on productivity. They assume that productivity increases result in proportionate
employment increases through shifts in labour demand. However, such inferences are dangerous because
others forces may be at work.
SECTION 2
7
This approach implicitly assumes that changes in labour supply are independent of local conditions. This
is a rather strong assumption: for example, congestion externalities such as higher rents and pollution,
are likely to influence mobility choices, potentially breaking the causality chain going from agglomeration economies to productivity and employment (Cingano and Schivardi, 2003:3).
The existence of migration between localities complicates the measurement of agglomeration effects. For
example, migration from poor to richer regions may eliminate differences in average incomes caused by
productivity differentials. Higher incomes in large cities may also be absorbed by higher property rents and
commodity prices (Glaeser, 2007). If they exist, higher incomes in large cities may reflect differences in
occupati­on (e.g. more managerial and professional jobs) rather than productivity per se. Glaeser (1993)
sugges­ts that growth in employment and population may be better indicators than average incomes.
The next sections present the evidence from developed and developing countries. Studies are classified
according to one of the three dimensions: industrial, geographic or temporal scope.
2.3.DEVELOPED COUNTRY EVIDENCE
2.3.1. Industrial scope
Rosenthal and Strange’s (2004) review of the evidence suggests that doubling city size increases productivity by somewhere between 3-8 per cent. The strength of urbanisation and localisation economies varies
across the studies considered. Henderson (1986) finds substantial evidence of localisation and almost no
evidence of urbanisation. Yet Glaeser et al (1992) finds that urbanisation economies are much more important for growth. Nakumura (1985) finds that both processes affect productivity. Below we consider evidence
of localisation and urbanisation economies separately.
2.3.1.1. Localisation economies
Glaeser et al (1992) consider the impact of industrial specialisation on employment growth for the six larges­t
industries in each of 170 US counties over the period 1956-1987. They find that increased specialisation
results in lower employment growth. Increasing the concentration of a given industry within a city by 10%
reduces employment growth by 12%.
Henderson (1995) also considers the link between specialisation and employment growth, using data for 8
industries: 3 high tech and 5 mature industries. Growth is examined for the period 1970-1987. The findings
are not robust: specialisation does not seem to affect employment growth for high tech industries, while it
has a positive effect on job growth in mature industries.
Cingano and Schivardi (2003) construct a measure of sectoral total factor productivity (TFP) using data from
the balance sheets of over 30,000 Italian firms. They find that specialisation has sizeable effects on TFP over
the period 1986-1998. Doubling the share of sectoral employment in a given location raises sectoral TFP by
0.2% a year, and increases the growth rate by 10%. Doubling the initial level of employment in manufacturing
raises TFP by 0.4% a year.
Glaeser et al (1992) find that increased competition (more firms) is positively associated with employment
growth, unlike specialisation. More firms per worker in a city-industry relative to the national average leads to
high growth of that city-industry: “Going from as many to twice as many firms per worker as the national averag­e
… raises growth of employment in the city-industry by 59 percent over 30 years” (Glaeser et al, 1992: 1144).
8
SECTION 2
2.3.1.2. Urbanisation economies
The main studies that measure urbanisation economies are Glaeser et al (1992), Henderson et al (1995) and
Rosenthal and Strange (2003). They all find that diversity fosters growth. Glaeser et al (1992) find that it
specifically supports employment growth; Henderson et al (1995) find that it promotes employment growth
among high-technology firms; and Rosenthal and Strange (2003) report a positive relationship with the
births of new firms. Duranton and Puga (2001) find that diverse cities in France encourage new industries to
emerge, which then move to specialised cities after they mature.
Graham (2007) uses company-level data to estimate the returns to agglomeration in the UK. He finds
positiv­e externalities for manufacturing, construction and six service industries. The estimated elasticities
range from 0.07 to 0.23, suggesting that doubling city size is associated with an increase in productivity
of between 7%-23%. Services seem to benefit more from agglomeration than manufacturing, particularly
transport, management consultancy, financial services and public services.
2.3.2.Geographical scope
Ciccone and Hall (1996) examine the geographical scope of agglomeration economies in the USA. They do
this by considering how county-level employment affects productivity at the wider state-level. They find that
doubling county employment density increases state productivity by around 6%. Ciccone (2002) uses similar
methods to estimate the effects of employment density on productivity for regions in France, Germany, Italy,
Spain and the UK. He finds the elasticity of productivity with respect to growth in Europe to be 4.5% – slightly
lower than the US result.
Rosenthal and Strange (2003) find that agglomeration economies diminish quite quickly with distance from
the core city. Soest et al (2002) concur that the benefits of agglomeration attenuate rapidly with distance,
even within cities.
Duranton and Overman (2002) consider both the industrial and geographical scope of agglomeration
economie­s and estimate the impact of density on productivity for 4 separate radii from the firm: 1km,
5km, 10km and 15km. They find positive localisation and urbanisation externalities for both manufacturing
and service industries. The weighted average localisation elasticity for manufacturing is 0.03 and 0.01 for
service­s. Furtherm­ore, localisation economies tend to exist over short distances (within 10km of firms).
A study in the UK by Rice et al (2006) extended this work by measuring the rate at which the advantages
of proximity diminish with travel time from the core city. They found that the benefits are greatest within
40 minutes driving time of the city core, tapering off quite sharply thereafter and having little or no effect
beyond about 80 minutes. The effects of agglomeration are four times stronger 30 minutes driving-time
away than 60 minutes away, and 17 times stronger than 90 minutes away.
2.4 DEVELOPING COUNTRY EVIDENCE
Most econometric studies of agglomeration have been in developed countries. There have been few studies
of agglomeration economies in developing countries, especially in Africa (Quigley, 2008). Table 2 summarise­s
this literature.
One message is that localisation economies are more prevalent in developing countries than urbanisation
economies. For example, Henderson (1988) finds that the concentration of firms in Brazil is positively related
to city growth. This is because “a clustered or densely populated region [provides] a rich environment for
competition and collaboration among firms and workers in the region, which lead to economic growth”
(Henderso­n, 1988: 23).
SECTION 2
9
TABLE 2: Analyses of agglomeration economies in developing countries
Country
Author (date)
Main conclusions
Brazil
Henderson (1988)
Localisation economies apparent
Korea
Henderson (2001)
Localisation economies in 3 industries. Urbanisation economies in 1 industry.
Lee and Zang (1998)
Localisation not urbanisation economies
China
Chen (1996)
Localisation economies
India
Shukla (1996)
Urbanisation stronger than localisation economies
Mitra (2000)
Urbanisation economies in 11 out of 17 industries
Lall et al (2003)
Urbanisation economies in 8 industries. Localisation diseconomies
Lall et al (2004)
No localisation or urbanisation economies
Indonesia
Henderson (1996)
Localisation economies in 3 industries. Urbanisation economies in 3 industries.
Source: Overman and Venables (2010)
Henderson et al (2001) estimate the relationship between agglomeration, output growth and employment
growth using industry data for South Korean cities. They find that localisation economies are positively
related to both output and employment growth. A 1% increase in local own industry employment results in a
0.06% - 0.08% increase in plant output. This is interpreted as follows:
a plant in a city with 1000 workers in other firms in the same industry would, without changing its own
inputs, increase its output by over 70% by moving to another city with 10000 workers in the same industry (Henderson, 2002: 92).
Lee and Zang (1998) estimate the effects of localisation and urbanisation economies on the productivity of 19
Korean manufacturing industries. Using cross-sectional census data for Korean cities the authors report that
localisation economies have been dominant for most manufacturing industries across Korea. Urbanisatio­n
economies were generally unimportant because of the negative effects of locating in large cities.
Cota (2001) looks at the effects of agglomeration on manufacturing employment in Mexico and finds
a positiv­e relationship between specialisation and job growth. A 10% increase in industry specialisation
boosts employment growth by 12%. This is the opposite of Glaeser’s (1992) finding. Pooled labour markets
in some of the northern Mexican cities could help to explain this positive effect.
Chen (1996) assesses the impact of agglomeration on productivity in the machinery and food industries
for 30 Chinese regions. He finds that agglomeration positively impacts productivity. However, the positive
impact diminishes with the growth of the city and then declines:
In Shanghai both industries [machinery and food] have surpassed the optimal agglomeration scale, and
the low efficiency firms have been squeezed out. Firm numbers in the machinery industry from 1987 to
1992 were 1522, 1641, 1696, 1661, 1623 and 1385; and firm numbers in the food industry, from 1988 to
1992, were 449, 447, 449, 442 and 347 (Chen, 1996: 429).
In research on India, Mitra (2000) also finds that the benefits of agglomeration diminish after a certain city
size threshold is reached. While TFP is generally responsive to urban population or industrial spread,
10
SECTION 2
productivity augmenting effects of urbanisation or urban industrial spread are not steady all through;
diseconomies outweigh the economies once urban population or urban manufacturing employment are
exceedingly large (Mitra, 2000:104).
Roberts and Goh (2010) assess whether density explains the spatial productivity disparities within Chongqin­g
in China. The estimated elasticity of productivity with respect to density is 3.6%.
Considering the temporal scope of agglomeration, Henderson and Kuncoro (1996) find that historic
concentratio­ns of particular employment sectors in Indonesia are linked with subsequent growth of the same
sectors. Examining five major capital goods industries (machinery, electrical machinery, primary metal­s,
transportation equipment and instruments), they find that patterns persist over time. Growth in traditio­nal
manufacturing jobs is higher in cities with high past concentrations of these industries, presumabl­y because
a conducive environment is created to attract further investment.
Fafchamps (2004) assesses the impact of agglomeration on employment and output growth for manufacturin­g
in Morocco, using firm-level census data. He concludes that agglomeration has a strong effect on both
measure­s of growth. However, the underlying mechanism is neither competition nor industrial diversity, in
contrast with the findings of Glaeser et al (1992).
Bigsten et al (2011) consider the impact of agglomeration in Ethiopia and find a positive, statistically significant relationship with productivity. They use census panel data for Ethiopian manufacturing firms and find
that for every additional firm producing the same product in a town, productivity rises by about 0.5%. This
supports the argument that localisation or clustering generates positive externalities. They suggest that trust,
cooperation and the informal enforcement of contracts between businesses is particularly important in countries with weak formal institutions. These social processes are much easier if firms are located close together.
To sum up, the broad message from previous research is that agglomeration tends to improve economic
performance. This is not universally true, and there are important differences in the nature and magnitude
of this effect depending on the local context and way in which density and growth are measured. There are
also some inconsistent and even contradictory findings between different studies.
2.5 EVIDENCE IN THE WORLD DEVELOPMENT REPORT
The 2009 World Development Report is largely a review, synthesis and repackaging of previous research.
It presents little original evidence to support a strong relationship between economic concentration and
growth. This is somewhat surprising considering the assertive tenor of the policy recommendations. There
are also few qualifications and conditions associated with these recommendations. Key authors of the report
subsequently explained that:
The WDR is parsimonious in its approach and limited in its scope. For a global report to be instructive
and informative, with findings and messages that are applicable to large and small countries at different
levels of development, it has to be stripped to its essence (Deichmann et al., 2011: 172).
One of the dangers in this approach is that it can oversimplify and exaggerate the contribution of cities to
economic development. Other important factors and forces that either create the conditions for agglomeration economies to take effect, or that may override the influence of agglomeration, are neglected. The connection between urbanisation and development ends up being portrayed as a kind of universal law. This is
inconsistent with the empirical evidence available.
SECTION 2
11
SECTION
3
ESTIMATION APPROACH
3.1. METHODS
Our empirical analysis relates measures of density to measures of growth at the local municipality level.
This is done using bivariate linear regression. Bivariate comparisons are useful indicators of underlying
relations­hips and one of the easiest ways to see whether a relationship exists (Wittenberg, 2010). This type
of correlatio­n cannot prove that density in 1996 caused growth over the period 1996-2010, but it can at the
least suggest that density and growth are strong complements. The regression model effectively treats all
other factors affecting growth as unobserved.
Using ordinary least squares (OLS) regression analysis, we estimate equations 1 and 2 as follows
Growth = a + b1 Log (Density) + e (1)
Growth = a + b2 (Skills) + u (2)
Equation 1 is a log-linear regression model. Both density and growth can be said to be in logarithmic form.
The logarithmic transformation of density is used in this equation and by definition, the exponential growth
model used to calculate population, employment and growth is in the logarithmic form1. One attractive
featur­e of this model, which has made it popular in applied work, is that the slope coefficient b1 can be
interpreted as an elasticity (Gujarati, 2003). In this case, b1 represents the elasticity of growth with respect
to density, that is, a percentage change in growth for a given (small) percentage change in density. Another
attractive feature of the model is that using the logarithms of the values rather than the actual values reduces
the wide range (particularly on the X axis) to a more manageable size.
In equation 2 only one variable (growth) appears in the logarithmic form. The equation can therefore be said
to represent a semi-log model. The slope coefficient measures the relative change in growth for an absolute
change in the proportion of the population with skills. Since the skills variable in the equation represents a
percentage (the proportion of the municipal population with matric), b2 can be interpreted as a percentage
change in growth given a percentage increase in the proportion of the population with skills. b2 is known as
a semi elasticity (Gujarati, 2003).
The variables e and u, called the error terms in the relationship, represent the ‘unobserved’ factors. The
intercept parameter a is not central to the analysis.
1
12
Exponential Growth= ln(xn/x1)/n, where n is the number of years in the period
If density contributes to growth, this should show up as a statistically significant relationship between at
least one set of density-growth regressions. Statistical significance allows us to infer that a relationship
between density and growth did not occur just by chance; rather that some fundamental relationship exists
between the variables. Significance is typically measured by the t-statistic (or p-value). We do not report the
t-statistic explicitly, but it can easily be calculated by dividing a coefficient by its respective standard error.
In all the regressions we report, the standard error is presented in parenthesis below the coefficient. If the
t-stat is more than 2 (the coefficient is at least twice as large as the standard error), we would generally
conclud­e that density has a significant impact on growth.
A series of scatter plots corresponding to the regression results is presented. These show the direction
and magnitude of the association between density in 1996 and economic growth over the subsequent
15 year period.
3.2.DATA
Data was obtained from two independent service providers, Global Insight and Quantec. In the absence
of official time-series data for localities in South Africa, these are the two main sources available. Having
data from two sources enables some cross-checking for consistency and reliability. After this process and
discussion­s with both providers we decided to rely on the Global Insight (GI) database, using the latest
versio­n of the IHS Regional Explorer (ReX) available at the time (March 2011). GI was chosen partly because
they have more staff available to update and check their data than Quantec. GI arrives at their sub-national
estimates by:
draw[ing] together many different sources of sub-national economic information from Statistics South
Africa, government departments, development agencies, Regional Services Councils, private research
houses and IHS Global Insight’s own data. These data components are reworked to ensure that they are
internally consistent and add up to the national totals (GI, 2010: 4).
A more detailed explanation of how GI achieves its estimates is presented in appendix A. It is fair to say that
the procedure remains unsatisfactorily vague in some respects.
The geographical units used are the 231 local municipalities and 6 metropolitan municipalities. South
Africa­n municipalities are relatively large compared with most countries, so they are more self-contained as
functional areas. This means that cross boundary flows of people and resources are smaller than in countries
with smaller municipal jurisdictions. Consequently, there is less leakage through commuting and trade, and
a stronger connection is likely between local population density, levels of economic activity and rates of
economic growth.
The boundaries used to define municipalities are based on the 2005 boundaries as reported by Stats SA.
The 2005 boundaries are chosen over the more up-to-date 2011 boundaries as the latter include several
sparse areas and national parks within the municipality boundary which would give a misleading impression
of density. The latest boundary changes also include the recent merger of Tshwane and Metsweding and the
reclassification of two new metros (Buffalo City and Mangaung). Since the period under consideration for the
analysis is 1996 to 2010, it seems sensible to use the original boundaries.
Population, employment, gross value added (GVA) and education data is obtained for each of the 237 municipalities and then manipulated to create the variables of interest for this study, namely measures of density,
skills share and growth.
SECTION 3
13
3.3. MEASURES
3.3.1. Dependent variables
Economic growth. The growth rate used in the analysis is based on continuous, exponential growth (or annual
compound growth) between two points in time. This is superior to the simple average growth rate as the
latte­r overstates the growth estimations because it neglects the fact that the base for growth is continual­ly
rising (Carlin, 2008). The three measures of growth are population, employment and GVA.
3.3.2. Independent variables
Density. Economic density is shorthand for the scale of output produced, and thus the income generated,
for a particular area. Gross value added is the key measure of economic output, based on the difference
between the value of goods and services produced and the cost of materials and other inputs. Gross value
added per km2 of land is used as one of the measures of density.
Knowledge spillovers, learning and labour market matching emphasize physical interaction as the way in
which information and ideas are spread. Our second and third measures of density therefore capture the
proximity between people/workers. One is an employment-based measure of density and the other is a
population-based measure (Abel et al, 2011).
Skills share. The conventional measure of human capital is based on education. It has been linked to several
measures of regional vitality (Abel et al, 2011). We use the share of people with matric to represent human
capital within a municipality. It does not capture the full rage of human capabilities and competences.
3.3.3. Categorisation
To convey an initial impression of the diverse composition of the 237 municipalities we disaggregate them
for descriptive purposes. Later we assess whether there is a stronger relationship between density and
growth in some groups than in others, in case the connection is diluted or washed-out by aggregating the
municipalities all together. Municipalities are put into rural and urban groups based on their historical
administrative classification and are also grouped according to their principal industry.
MAP 1. The Distribution of
municipalities by historical
administrative classification, 1996
14
SECTION 3
Rural and urban groups
This paper defines rural and urban municipalities in terms
of apartheid categories and in line with Makgetla’s (2010)
approach. Rural areas cover the former Bantustans and
commercial farming areas. Urban areas cover the metros
and secondary cities. The distribution of municipalities by
historical administrative classification is shown in Map 1
(see p16).
35
33
30
25
22
20
17
15
10
5
5
5
1
0
KZN
EC
LIM
MP
NW
NC
Rural Municipalities
Former Bantustans. Under apartheid Africans who made
up 80% of the population of South Africa were restricted
to land ownership in ‘rural reserves’, ‘homelands’ or
‘forme­r Bantustans’ (Baldwin, 1975). They were typically
located on the periphery of and distant from the main
economic centres. These areas were typically arid with
very limited agricultural and mining potential.
GRAPH 1: Number of municipalities in the former Bantustans
per province
30
25
25
21
20
With few exceptions, the Bantustan administra-
15
tions had virtually no resource base of their own,
10
and the central state provided only limited subsi-
5
dies. The Bantustans ended up with too few and
0
often poorly qualified educators, police and health
workers. They suffered from severe underinvestment in both economic and household infrastructure, leaving them with inadequate transport, com-
18
15
14
12
9
NC
WC
FS
EC
KZN
NW
MP
7
6
LIM
GT
GRAPH 2: Number of municipalities in commercial farming areas
per province
munications, power and irrigation for producers
as well as enormous backlogs in residential water,
sewage and electricity (Makgetla, 2010: 19).
The bulk of the municipalities located in the former Bantustans are contained in the Eastern Cape, Limpopo and
Kwazulu Natal. A few are also situated in Mpumalanga,
the North West and the Northern Cape provinces.
5
4
4
4
3
3
3
2
2
Commercial farming areas. These areas are made up of
smaller towns, commercial farms and most mining areas
(Makgetla, 2010). The Western Cape, Northern Cape and
the Free State contain extensive commercial farming
regions but almost no former Bantustan areas. Graph 2
shows the number Commercial farming areas by province.
2
1
1
1
EC
LIM
NC
1
0
MP
NW
KZN
WC
FS
GT
GRAPH 3. Number of secondary city municipalities per province
Urban municipalities
Secondary cities. These areas tend to have narrow economic bases and specialised industries. Hence they
face diverse processes of growth and decline, depending on the performance of their basic industries. The
distribution of the municipalities of secondary cities by province can be seen in Graph 3.
Metropolitan municipalities. The 6 metros of Cape Town, Nelson Mandela Bay, eThekwini, Johannesburg,
Tshwane and Ekurhuleni are the largest agglomerations and serve as the economic engines for their
SECTION 3
15
surrounding areas. They have most potential to address the socio-economic needs of the population. They
have diversified economies, large-scale productive infrastructure, and produce a variety of goods and services for national and international markets.
Built-up area. An additional set of results is presented for the 27 urban municipalities. This is based on
redefining the area of analysis as the built-up area. Sparsely populated areas within each municipality are
deliberately excluded. This procedure is particularly important for the metros because they have generous
boundaries including substantial undeveloped areas. We define built-up area as the land area covered by
those sub-places with a population density above 25 people per hectare or 2500 people per km2. This is
lower than in many countries but consistent with South Africa’s distinctive urban form comprising concentrated populations on the urban outskirts. Low density middle and high income suburbs and high density
townships are a legacy of separate development under apartheid.
We assume areas with a population density lower than 25 people per ha or 2500 people per km2 are either
sparsely populated or have low levels of economic activity. These areas do not form part of the city’s core and
are therefore excluded from the built-up area.
Built-up area density is calculated as follows
1.
Summing the area of all sub places within the municipality with a population density higher
than 2500 people per km2.
2.
Summing the 2001 Census population and employment totals of all the sub-places that lie
within the defined built up area of a municipality.
3.
Deriving the population and employment densities by dividing the population and employment totals of the built up area by the built up area.
Table 3 shows the total number of municipalities in each province and their average population. Gauteng
has by far the fewest municipalities and the biggest in terms of average population. At the other end of the
spectrum, Northern Cape has many municipalities with small populations, because of its physical extent and
sparcity. The provinces with most former Bantustans (Limpopo, Kwazulu Natal and the Eastern Cape) have
relatively large average populations.
TABLE 3: Number of municipalities and average population per province
16
SECTION 3
Province
Number of municipalities
Ave. population of each municipality
Gauteng
11
748000
Limpopo
25
186500
Mpumalanga
18
174600
Kwazulu Natal
51
173200
Western Cape
25
161700
Eastern Cape
39
158500
North West
21
143800
Free State
20
133000
Northern Cape
27
35700
South Africa
237
176100
Industrial composition
Municipalities are also categorised by their dominant tradable sector to assess whether, for example,
agglomeration economies are stronger in manufacturing than in agricultural areas. The procedure followed
ensured that the final groups were mutually exclusive:
1.
Calculate the proportion of employment in each broad tradable sector for all municipalities
in 1996.
2.
Allocate each municipality to the category agriculture, mining or manufacturing according to
its largest sector.
3.
If none of these sectors employed more than 10% of the workforce, that municipality was
not considered to have a dominant tradable sector and was categorised as community (i.e.
public) services (which includes health, education and social services)
Using this procedure, we identify:
146 agricultural municipalities, with a share of jobs in agriculture ranging from 10-65%.
25 mining municipalities, with the share of mining jobs ranging from 15-75%.
43 manufacturing municipalities, with the share of manufacturing jobs between 13-36%.
23 community services municipalities, with these jobs ranging from 30-57%.
The breakdown of municipalities in each province by local economic structure is shown in Table 4. Their
national distribution is shown in Map 2. The majority of municipalities are agricultural in all provinces except
Gauteng, where most are manufacturing.
TABLE 4: Number of municipalities in each province by industrial composition
Province
Manufacturing
Agriculture
Mining
Community services
Gauteng
9
0
2
0
Mpumalanga
1
11
4
2
North West
1
14
5
1
Free State
3
15
2
0
Limpopo
0
16
4
5
Western Cape
7
18
0
0
Northern Cape
0
19
7
1
Eastern Cape
8
23
0
8
Kwazulu Natal
14
30
1
6
South Africa
43
146
25
23
SECTION 3
17
MAP 2: The Distribution
of Municipalities by
Industrial Composition
18
SECTION 3
SECTION
4
EMPIRICAL ANALYSIS: Descriptive
4.1. THE DISTRIBUTION OF DENSITY IN SOUTH AFRICA
There is a hierarchy of density across municipalities. Primary cities are at the top and rural farming areas are
at the bottom. Table 5 shows the 10 municipalities with the lowest and highest population densities. The top
10 are metros or secondary cities, and are predominantly manufacturing. Four are in Gauteng. The bottom 10
are rural and historically categorised as commercial farming. Nine are predominantly agricultural and one is
mining. Eight are in the Northern Cape. Laingsburg is in the semi-desert area of the Karoo and Molopo is in
North West province. Both municipalities border on the Northern Cape.
TABLE 5: The ten highest and lowest ranked municipalities by population density
Rank
Municipality
Administrative classification
Sector
Province People/km2
1
2
Johannesburg
eThekwini
Urban
Urban
Metro
Metro
Manufacturing
Manufacturing
3
Ekurhuleni
Urban
Metro
Manufacturing
GT
1137
4
5
6
7
Cape Town
Msunduzi
Tshwane
Emfuleni
Urban
Urban
Urban
Urban
Metro
Secondary
Metro
Secondary
Manufacturing
Community
Community
Manufacturing
WC
KZN
GT
GT
1063
854
841
672
8
N Mandela Bay
Urban
Metro
Manufacturing
EC
497
9
10
Buffalo City
uMhlathuze
Urban
Urban
Secondary
Secondary
Manufacturing
Manufacturing
EC
KZN
272
251
228
229
Khai-Ma
Molopo
Rural
Rural
Commercial
Commercial
Agriculture
Agriculture
NC
NW
1
1
230
Siyathemba
Rural
Commercial
Agriculture
NC
0.9
GT
KZN
1715
1229
231
Ubuntu
Rural
Commercial
Agriculture
NC
0.9
232
Kamiesberg
Rural
Commercial
mining
NC
0.8
233
234
235
Laingsburg
Kareeberg
Mier
Rural
Rural
Rural
Commercial
Commercial
Commercial
Agriculture
Agriculture
Agriculture
WC
NC
NC
0.6
0.6
0.5
236
237
Karoo Hoogland
Hantam
Rural
Rural
Commercial
Commercial
Agriculture
Agriculture
NC
NC
0.4
0.3
19
Map 3 shows the national distribution of population density. The key on each map is structured such that the
bottom 10 ranked municipalities are in the lowest category and the highest 10 ranked municipalities are in
the highest category. The most extensive areas of high density are in Kwazulu Natal, Eastern Cape, Gauteng
and Limpopo.
MAP 3. Population density by
local municipality, 1996
Table 6 shows the pattern for employment density rather than population. The result is very similar. The
metro­s have the highest job densities followed by two gold mining municipalities close to Gauteng (Randfontein and Westonaria). Municipalities with the lowest job densities are generally commercial farming areas
in the Northern Cape
TABLE 6: The ten highest and lowest ranked municipalities by employment density
20
SECTION 4
Rank
Municipality
Administrative classification
Sector
Province
Jobs/km2
1
2
Johannesburg
Cape Town
Urban
Urban
Metro
Metro
Manufacturing
Manufacturing
GT
WC
787
349
3
Ekurhuleni
Urban
Metro
Manufacturing
GT
322
4
5
6
Tshwane
Urban
Metro
Community
GT
312
eThekwini
Msunduzi
Urban
Urban
Metro
Secondary
Manufacturing
Community
KZN
KZN
307
165
7
N Mandela Bay
Urban
Metro
Manufacturing
EC
108
8
Emfuleni
Urban
secondary
Manufacturing
GT
105
9
Westonaria
Urban
Commercial
Mining
GT
70
10
Randfontein
Urban
Commercial
Mining
GT
66
228
Siyancuma
Rural
Commercial
Agriculture
NC
0.2
229
Kamiesberg
Rural
Commercial
Mining
NC
0.2
230
231
Siyathemba
Ubuntu
Rural
Rural
Commercial
Commercial
Agriculture
Agriculture
NC
NC
0.2
0.2
232
Laingsburg
Rural
Commercial
Agriculture
WC
0.2
233
Kareeberg
Rural
Commercial
Agriculture
NC
0.1
234
Karoo Hoogland
Rural
Commercial
Agriculture
NC
0.1
235
Mier
Rural
Commercial
Agriculture
NC
0.1
236
Molopo
Rural
Commercial
Agriculture
NW
0.09
237
Hantam
Rural
Commercial
Agriculture
NC
0.08
Map 4 shows the national distribution of employment density. The areas of high job density are limited to the
metros and a few secondary cities. Elsewhere, job densities tend to be low. The contrast between the areas
of high population density in Kwazulu Natal, Eastern Cape and Limpopo and their low employment densities
is very striking. There is clearly a serious jobs shortfall in these places.
MAP 4. Employment density by
local municipality, 1996
The map of GVA density (Map 5) is slightly different again, partly because mining and tourism areas feature
more prominently than they do on the employment density map.
MAP 5. GVA density by local
municipality, 1996
SECTION 4
21
Table 7 shows the distribution of people with matric. The 10 municipalities with the lowest level of skills are
all in the former Bantustans, reflecting the low investments in African education under Apartheid (Banjeree
et al, 2007). The 10 municipalities with the highest skills are a mixture of cities, towns and rural areas. The
pattern is quite different to the density hierarchy.
TABLE 7: The ten highest and lowest ranked municipalities by skills share
Rank
Municipality
Administrative classification
Sector
Province
Matric (%)
1
Tshwane
Urban
Metro
Community
GT
18.7
2
Midvaal
Rural
Commercial
Manufacturing
GT
17.9
3
Johannesburg
Urban
Metro
Manufacturing
GT
17.3
4
Stellenbosch
Urban
Secondary
Agriculture
WC
16.6
5
Mogale City
Urban
Secondary
Community
GT
16.5
6
Mookgopong
Rural
Commercial
Agriculture
LIM
15.9
7
Ekurhuleni
Urban
Metro
Manufacturing
GT
15.4
8
Overstrand
Rural
Commercial
Agriculture
WC
15.2
9
Tlokwe
Urban
Secondary
Agriculture
NW
15.1
10
NokengtsaTaemane
Rural
Commercial
Mining
GT
14.8
Ratlou
Matatiele
Rural
Rural
Former Bantustan
Former Bantustan
Agriculture
Agriculture
EC
KZN
2.3
2.3
230
Emalahleni
Rural
Former Bantustan
Agriculture
EC
2.3
231
uMuziwabantu
Rural
Former Bantustan
Agriculture
KZN
2.3
232
Ratlou
Rural
Former Bantustan
Agriculture
NW
2.3
233
Elundini
Rural
Former Bantustan
Agriculture
EC
2.1
228
229
234
Mbhashe
Rural
Former Bantustan
Agriculture
EC
1.9
235
Port St Johns
Rural
Former Bantustan
Agriculture
EC
1.8
236
Instika Yethu
Rural
Former Bantustan
Agriculture
EC
1.7
237
Ntabankulu
Rural
Former Bantustan
Agriculture
EC
1.5
Map 6 shows the national distribution of skills. It confirms the stronger position of the cities.
MAP 6. Share of the population
with matric by local municipality,
1996
22
SECTION 4
Table 8 compares the higher and lower ends of the population density distribution for the former Bantustans
and commercial farming areas. There are a few high density commercial farming areas that include towns
and tourism destinations. Otherwise their densities are generally lower than the former Bantustans. This is
also clear at the lower end of the density distribution.
TABLE 8: The ten highest and lowest ranked rural municipalities by population density
Rank
Former Bantustans
People/km2 Commercial farming areas
People/km2
1
Bushbuckridge
212
Randfontein
243
2
Mandeni
198
Hibiscus Coast
238
3
Thulamela
187
Umdoni
233
4
Dr JS Moroka
183
KwaDukuza
211
5
Ndwedwe
156
Westonaria
189
6
Imbabazane
140
Merafong City
131
7
Umzumbe
138
Nquthu
80
8
Maphumulo
138
Maluti a Phofung
80
9
King Sabata Dalindyebo
131
Mafikeng
66
10
Makhuduthamaga
130
Phokwane
66
228
Maruleng
27
Khai-Ma
1.1
229
Elundini
26
Molopo
1.1
230
Sakhisizwe
26
Siyathemba
0.9
231
The Big Five False Bay
25
Ubuntu
0.9
232
Great Kei
23
Kamiesberg
0.8
233
Ratlou (Setla-Kgobi)
21
Laingsburg
0.6
234
Senqu
17
Kareeberg
0.6
235
Moshaweng
10
Mier
0.5
236
Kagisano
7
Karoo Hoogland
0.4
237
Tsolwana
5
Hantam
0.3
Table 9 (see p26) compares the higher and lower ends of the employment density distribution for the former
Bantustan­s and commercial farming areas. The former are much worse off in terms of the ratio of employment to population. For example, Bushbuckridge has the highest population density among the former
Bantustans but only 12 jobs per km2, i.e. 1 job for about 18 people. Randfontein is the commercial farming
municipality with the highest population density and has 66 jobs per km2, i.e. about 1 job for every 4 people.
Graphs 4 and 5 (see p26) show the spread or distribution of the population and employment densities for
the former Bantustans and the commercial farming areas. There are a few places with high population and
employment densities among the commercial farming areas, but a more even distribution among the former
Bantustan­s. Bearing in mind the different scales of the two figures (especially for employment density), the
other importan­t point emerging is that the ratio of population to employment in the former Bantustans is
generally much higher than in the commercial farming areas.
SECTION 4
23
TABLE 9: The ten highest and lowest ranked rural municipalities by employment density
Rank
Former Bantustans
Jobs/km2
1
Mandeni
24
Westonaria
71
2
Mbonambi
16
Randfontein
66
3
Greater Tzaneen
13
Merafong City
59
4
Thulamela
13
KwaDukuza
51
5
Umzumbe
13
Hibiscus Coast
41
6
Greater Letaba
13
Umdoni
22
7
Imbabazane
12
Mafikeng
16
8
King Sabata Dalindyebo
12
Umjindi
16
9
Bushbuckridge
12
Metsimaholo
15
10
uMlalazi
11
Breede Valley
14
228
Elundini
1.6
Siyancuma
0.2
229
Emalahleni
1.5
Kamiesberg
0.2
230
Kagisano
1.4
Siyathemba
0.2
231
Moshaweng
1.4
Ubuntu
0.2
232
Mutale
1.3
Laingsburg
0.2
233
Okhahlamba
1.1
Kareeberg
0.1
234
Senqu
Karoo Hoogland
0.1
235
The Big Five False Bay
0.8
Mier
0.1
236
Tsolwana
0.5
Molopo
0.09
237
Umhlabuyalingana
0.5
Hantam
0.08
1
Population density
Commercial farming areas
Jobs/km2
Employment density
Number of people/sq km
Number of people employed/sq km
GRAPH 4: Population versus employment density. Commercial farming areas 1996
Population density Employment density
Number of people/sq km
Number of people employed/sq km
GRAPH 5: Population and employment density. Former Bantustans 1996
24
SECTION 4
SECTION
5
EMPIRICAL ANALYSIS: Summary statistics
Before embarking on a systematic analysis of the relationship between density and growth, it is instructive
to explore summary statistics for the different types of municipality.
5.1. NATIONAL, URBAN AND RURAL MUNICIPALITIES
Table 10 presents summary statistics for rural and urban municipalities in 1996. The mean population density
in urban municipalities was about nine times higher than in rural areas. The mean employment density in
urban areas was about 20 times higher than in rural areas. And the mean GVA density in urban areas was
nearly 30 times the rural figure. This implies that urban areas were considerably more productive than rural
areas. This is defined in conventional economic terms, which excludes informal economic activity.
TABLE 10: National, urban, and rural summary statistics
National
Urban
Rural
Mean population density
83
381
43
Mean employment density2
18
114
6
Population : employment ratio3
5
3
7
R2.1m
R14.2m
R0.5m
Mean share of skills (%)
7.1
13
6.3
Mean population growth (%)
1.1
1.4
1
Mean employment growth (%)
1.4
2.2
1.3
Mean GVA growth (%)
2.1
2.6
2.1
1
Mean GVA density
4
5
Notes: 1 = Number of people per km2 2 = Number of people employed per km2
3 = Number of people/number of jobs 4 = Output per km2 5 = Share of the population with matric
In addition, there were twice as many people per job in the rural than in the urban areas. This gives an indication of the extra pressure on the productive population in rural areas. And the proportion of people with matric in urban areas was twice as high as that in rural areas. Urban municipalities experienced stronger growth
than rural areas between 1996-2010 across all three indicators, but particularly in terms of employment.
25
5.2.BUILT-UP URBAN AREAS
In section 3.3.3 it was noted that many urban municipalities have generous boundaries, so the analysis
of density was refined to focus on the built-up area of the 27 urban municipalities. Table 11 presents the
descriptive statistics for these areas against the total land area of the same municipalities. This procedure
clearly makes a big difference to all density measures.
TABLE 11: Built-up area summary statistics
Urban built-up
Urban
Mean population density
5630
381
Mean employment density2
1337
114
R342m4
R14.2m
1
Mean GVA density3
Notes: 1 = Number of people per km2 2 = Number of people employed per km2
3 = Output per km2 4 = We assume total municipal GVA is produced on the built up area
5.3. HISTORICAL AND ADMINISTRATIVE CLASSIFICATION
In order to test for the influence of institutional and political factors on economic performance, we subdivided rural areas into the former Bantustans and commercial farming areas. We also sub-divided urban
areas into metros and secondary cities, since the metros have distinctive powers and responsibilities.
Table 12 presents summary statistics for these four groups of municipalities. The metros have about six
times higher population densities and eight times higher employment densities than secondary cities. The
population:jobs ratio is more similar. The metros are best placed of all areas in terms of the key economic
indicators: GVA density, GVA growth and employment growth.
TABLE 12: Summary statistics for municipalities by historical administrative classification
Metros
Secondary
Cities
Commercial
farming areas
Former
Bantustans
Mean population density1
1081
182
25
72
Mean employment density2
365
42
6
5
3
4
4
15
R46.1m
R5.1m
R0.5m
R0.6m
Skills share (%)
15
12
8
4
Mean population growth (%)
1.3
1.3
1.4
0.4
Mean employment growth (%)
2.3
2.1
1.7
0.7
Mean GVA growth (%)
3.4
2.4
2.3
1.8
Mean population:employment3
Mean GVA density
4
5
Notes: 1 = Number of people per km2 2 = Number of people employed per km2
3 = Number of people/number of jobs 4 = Output per km2 5 = Share of the population with matric
26
SECTION 5
Comparing the population densities of the former Bantustans and commercial farming areas is very striking. The former is nearly three times higher than the latter. This is not matched by equivalent employment
differences. Consequently, there are nearly four times as many people per job in the former Bantustans as
in the commercial farming areas, with all the attendant pressures associated with this. With slower growth
in employment in these areas, it is perhaps not surprising that population growth has lagged substantially
behind other parts of the country, presumably because of net out-migration.
5.4.INDUSTRIAL COMPOSITION
Different industries tend to grow at different rates over time, and this may outweigh the effects of densit­y
or agglomeration. Table 13 presents descriptive statistics of the position of different municipalities grouped
according to their dominant economic sector. Manufacturing areas have much higher population and
employme­nt densities than other areas. They have also had higher GVA and employment growth rates.
Municipalities dominated by community services have the highest ratio of population to employment of all
areas, suggesting a big shortfall in jobs and weak local economies, perhaps compensated to a small extent
by public sector employment. Mining areas have not performed well over the last 15 years.
TABLE 13: Summary statistics for municipalities grouped by industrial composition
Manufacturing
Agriculture
Mining
Mean population density1
262
32
56
Mean employment density2
69
5
16
Mean population:employment3
4
6
4
R8.6m
R0.5m
1.9m
Mean Skills share (%)5
9.9
6.3
8.4
Mean population growth (%)
1.2
1.2
0.9
Mean employment growth (%)
1.8
1.5
0.3
Mean GVA growth (%)
2.8
2.2
1.0
Mean GVA density4
Community
services
95
6
15
0.6m
5
0.4
1.3
1.7
Notes: 1 = Number of people per km2 2 = Number of people employed per km2
3 = Number of people/number of jobs 4 = Output per km2 5 = Share of the population with matric
SECTION 5
27
SECTION
6
EMPIRICAL ANALYSIS: Regression results
6.1. INTRODUCTION
The purpose of the regression analysis is to establish how growth relates to density. We do this by estimatin­g
the average value of growth over the period 1996-2010 in terms of density in 1996. Regression implies
(without proving) causality between the independent (or explanatory) variable density and the dependent
variabl­e growth. Correlation implies no causality but refers simply to the degree of association between two
variables. For example, density and growth may be correlated because another variable affects them both.
We present statistically significant regressions and the graphical correlations for some of them. Different
combinations of density and growth measures are presented to illustrate the range of analyses that were
undertaken. GVA is the key growth measure in the analysis. The other growth measures produce similar
results. The choice of which growth and density indicators to present was influenced by the desire to convey
a rounded picture of all the evidence.
6.2.DENSITY RESULTS
The results obtained using different measures of density and growth are consistent. The basic message is
that when the analysis covers all 237 municipalities there is no relationship between density and growth.
This finding is surprising considering the body of evidence from elsewhere in the world. It suggests that
spatial proximity does not inevitably create positive externalities that raise productivity and increase growth.
This may be because other factors play a more important role in shaping local growth and development, and
perhaps even in masking the effects of agglomeration in some instances. These could include the industrial
structure, institutional arrangements, inherited conditions or macro-economic circumstances.
Before leaping to conclusions, another important finding is that, when separate regressions are run on different kinds of municipality, there is at least one significant relationship between density and growth in each
of the following categories: ‘metros’, ‘commercial farming areas’ and ‘community services areas’.
6.2.1.National, Urban, Rural and Built-up Urban area
6.2.1.1. National, Urban and Rural
Appendix C presents the regressions run on all 237 municipalities and on the urban and rural categories.
There is no significant relationship between density and growth for any of these groups. Graph 6 shows the
28
simple correlation between population density and GVA growth for all 237 municipalities, without using
the log scale to begin with. It distinguishes between rural areas (red dots) and urban areas (green dots).
Although there is a slight upward gradient on the line of best fit between the observations, its gradual slope
and the wide scatter of observations indicate that density and growth are not strongly related. Areas with the
highest densities have grown no faster than many areas with much lower densities. This finding holds when
any measure of density is regressed against any measure of growth.
In graph 6 it is difficult to see the distribution of areas in the 0 to 200 people per km2 range clearly. The
remaining graphs in this section therefore use the log scale instead. This also allows us to interpret the slope
coefficient as an elasticity. Graph 7 presents the same correlation, but on a log scale. This spreads out the distribution along the horizontal axis, and confirms the absence of a relationship between density and growth.
GRAPH 6: Population density versus economic growth (GVA)
Mossel Bay
Bela−Bela
2
6
y=2.06 + 0.0007x, R =0.009
George
Growth in GVA (%)
0
2
4
Tshwane
N Mandela
Stellenbosch
Johannesburg
eThekwini
Ekurhuleni
Emfuleni
−2
Matjhabeng
−4
Matlosana
0
500
1000
2
Number of people per km
Rural
1500
2000
Urban
GRAPH 7: Population density versus economic growth (GVA)
Bela−Bela
6
Bitou
Growth in GVA (%)
0
2
4
Tshwane
Joburg
Hibiscus Coast
Cape Town
Ekurhuleni
Umdoni
Randfontein
Emfuleni
−2
Matjhabeng
Matlosana
Merafong City
Westonaria
−4
Dannhauser
0
2
4
2
Log(Number of people per km )
Rural
2
y=2.1155 + 0.0024x, R =0.00
Mossel Bay
6
8
Urban
SECTION 6
29
There are about a dozen rural areas that have grown faster than all the urban areas, led by Bela-Bela
(Limpop­o), Mossel Bay and Bitou (both in the Western Cape). All three have strong tourism sectors and wellconnected transport routes. Bela-bela is located off the N1 highway between Tshwane and Polokwane and
is known for its mineral springs (warmbaths). Mossel Bay and Bitou (formerly Plettenburg Bay) are on the
south-east coast (the Garden route) and served by the N2 coastal highway.
Urban areas obviously have a higher population density than rural, although there are several exceptions. The
relatively high density rural areas of Hibiscus Coast, Umdoni and Randfontein are served by major national or
provincial roads. Hibiscus Coast and Umdoni are situated in the Ugu district municipality on the KZN South
Coast, about 150km south of eThekwini on the N2 highway. It is part rural and part urban, and houses some
major industrial complexes. Hibiscus Coast is the most concentrated economic hub in Ugu district. Randfontein is a gold mining town about 40km West of Johannesburg and is served by the N12 and R41 highways.
The rural areas of Merafong City (North West), Westonaria (Gauteng) and Dannhauser (Kwazulu Natal) are
some of the municipalities that have experienced declining GVA. Merafong City and Westonaria are on the
West Rand where gold mining is the principal activity and Dannhauser (located halfway between eThekwini
and Johannesburg) is surrounded by some of the largest coal mines in KZN. The decline in GVA may be linked
with shrinking mining activities. The urban areas of Matjhabeng (Free State) and Matlosana (North West) are
also experiencing declining GVA. Both are administrative districts.
The correlation between employment density and employment growth in Graph 8 yields similar results.
There is no significant relationship between these variables. The growth leaders and laggards consist of a
few select rural areas. The municipalities of Mthonjaneni and Ntabankulu experienced relatively high jobs
growth over the period, but are both among the poorest 5% of municipalities in South Africa. Ntabankulu
is the poorest municipality with 85% of its residents living below the poverty line (Schwabe, 2004). The
relativel­y high jobs growth in these places was probably linked to their low starting point in 1996.
2
y=0.0298 + 0.1548x, R =0.0042
Mthonjaneni
Ntabankulu
Bitou
Mkhambathini
uMhlathuze
Rustenburg
Johannesburg
Matjhabeng
Richmond
−10
Growth in employment (%)
−5
0
5
10
15
GRAPH 8: Employment density versus employment growth
Emfuleni
Westonaria
Dannhauser
0
2
4
6
2
Log(Number of people employed per km )
Rural
8
Urban
6.2.1.2. Built-up urban areas
Appendix D presents the results for the regressions focused on the built-up area of the urban municipalities.
No statistically significant relationship was found between density of the built-up area and the subsequent
30
SECTION 6
growth of the municipal area as a whole. Graphs 9 and 10 illustrate two of the correlations, involving population density and employment density. Statistical ‘insignificance’ aside, it is notable that the relationship
between population density and GVA growth is marginally negative, while that between employment density
and GVA growth is the opposite, perhaps hinting that employment density may have more influence on GVA
growth than population density.
The former industrial city of George (Western Cape) has the highest GVA growth rate among urban municipalities. It is situated on the Garden route, promotes itself as a ‘tourism mecca’ and is regarded as the administrative capital of the Southern Cape. Matlosana (North West) and Matjhabeng (Free State) have the lowest
GVA growth rates, perhaps because they are located in largely rural provinces. Drakenstein (Western Cape)
has the highest population density and is situated in the Cape Winelands district municipality. It is made up
of a concentration of towns and farming communities. Mbombela (Mpumalanga) covers an extensive area
including Nelspruit, and a mixture of agricultural, commercial and manufacturing activities.
GRAPH 9: Population density versus economic growth (GVA)
2
6
y = 8.8643 − 0.7234x, R = 0.014
GVA growth (%)
0
2
4
Polokwane
George
Tshwane
Rustenburg
Joburg
Cape Town
Drakenstein
Mbombela
Mogale City
Stellenbosch
Emfuleni
−2
Matjhabeng
Matlosana
8
8.2
8.4
8.6
2
Log(Number of people per km )
Secondary
8.8
9
Metro
GRAPH 10: Employment density versus economic growth (GVA)
2
6
y = 0.3122 + 0.4130x, R = 0.0089
9
GVA growth (%)
0
2
4
George
Polokwane Tshwane
Mbombela
Newcastle
Rustenburg
Joburg
eThekwini
Cape Town
Drakenstein
Mogale City
Emfuleni
Stellenbosch
−2
Matjhabeng
Matlosana
6
6.5
7
2
Log(Number of people employed per km )
Secondary
7.5
Metro
SECTION 6
31
6.2.1.3. Summary
This section considered the relationship between density and growth across four groups of municipality:
national, urban, rural and built-up urban area. The essential finding is that there is no apparent relationship
between density and growth. Municipalities with higher population or economic densities have grown no
faster than municipalities with lower densities.
6.2.2.Historical administrative classification
It has been established that there is no relationship between density and growth when all 237 municipalities
are included. Graph 11 shows this again, while distinguishing between former Bantustans (red dots), commercial farming areas (green dots), secondary cities (orange dots) and metros (blue dots). This highlights the relatively low employment density of the rural areas. Beyond a pronounced threshold, illustrated by the red line,
there lies only one former Bantustan and six commercial farming areas. The actual value represented by this
line is 22 people employed per km2 of land. The maximum employment density is in Johannesburg, with 787
jobs per km2. This is 10 times the highest employment density for a rural area (Westonaria, also in Gauteng).
GRAPH 11: Employment density versus economic growth (GVA)
2
y = 1.98 + 0.1039x, R = 0.0113
Nokeng
Mossel Bay
Bitou
Knysna
Growth in GVA (%)
0
2
4
6
Bela−Bela
George
Rustenburg
KwaDukuza
Hibiscus Coast
Mandeni
Umdoni
Tshwane
Johannesburg
Cape Town
Ekurhuleni
Randfontein
−2
Vulamehlo
−4
Merafong City
Dannhauser
0
Westonaria
2
4
6
2
Log(Number of people employed per km )
Former Bantustan
Secondary
8
Commercial
Metro
Appendix E presents the regressions run on each of the four sub-groups based on the historical administrative classification: former Bantustans, commercial farming areas, secondary cities and metros. This allows
for some control over the country’s unusual economic geography as a result of apartheid spatial engineerin­g.
Statistically significant results were obtained in the ‘metros’ and ‘commercial farming’ categories.
6.2.2.1. Metro results
The association between population growth and population density in the six metros is positive and
statistica­lly significant at the 10% level. This may be the result of higher natural growth rates in these areas,
given their relatively young populations, and/or higher net migration to the metros because they seem to
offer bette­r employment opportunities and greater access to education and health amenities. Interpreting
the coefficient of density as an elasticity, we find that, holding all else constant, a 1% increase in population
densit­y increases population growth by about 0.4%. Or a 10% increase in population density increases population growth by about 4%. Standard errors are in parentheses2. Note that there is no statistically significa­nt
relationship between any measure of density and economic growth in the metros.
Population growth = -1.2269 + 0.3653*(Log(Population density)), N=6, R2=0.5726
(1.0937) (0.1578)
32
SECTION 6
(1)
6.2.2.2. Commercial farming area results
For the 127 commercial farming areas, employment growth is positively related to all three density measure­s.
These relationships are all statistically significant at the 1% level. Holding all else constant, increasing any
one of the three measures of density by 1% increases employment growth by about 0.4%. Or increasing
densit­y in commercial farming areas by 10% increases the rate of employment growth by about 4%.
Employment growth=0.9019 + 0.36*(Log(Population density)), N=83, R2=0.0645
(2)
(0.3325) (0.1227)
Employment growth= 1.4310 + 0.3881*(Log(Employment density)), N=83, R2=0.0775
(3)
(0.1921) (0.1198)
Employment growth=-2.7084 + 0.3676*(Log(GVA density)), N=83, R2=0.0769
(4)
(1.3652) (0.1118)
Graph 12 shows the relationship between employment growth and employment density for the commercial
farming areas. The relationship is positive and statistically significant. Areas with higher concentrations of
employment (towns) are growing faster than villages and dispersed rural areas. Several areas have experienced employment decline, including Randfontein, Merafong City and Westonaria. All three are located in
Western Gauteng or on the West Rand. Gold mining is the main activity and the job losses in these areas may
be attributable to mining contraction.
10
GRAPH 12: Employment density versus employment growth
2
y = 1.4310 + 0.3881x, R =0.0775
Growth in employment (%)
0
5
Bitou
Greater Kokstad
Kungwini
Umdoni
KwaDukuza
Hibiscus Coast
Randfontein
Magareng
Merafong City
−5
Westonaria
0
1
2
3
2
Log(Number of people employed per km )
4
Commercial farming area
6.2.2.3. Summary
This section considered the relationship between density and growth across the four historical categories
of municipality. Statistically significant relationships were found between density and growth in the metros
and commercial farming areas. The latter seems more important because all three measures of density were
linked with subsequent employment growth.
SECTION 6
33
6.2.3.Local Economic Structure
This section disaggregates the municipalities by industrial structure to assess whether the relationship
between density and growth is influenced by the composition of the local economy. Each municipality is
categorised into one of four subgroups: agriculture, mining, manufacturing and community services.
Graph 13 shows the correlation between population density and subsequent population growth for all 237
municipalities. The graph distinguishes between agricultural areas (green dots), mining areas (red dots),
community services (orange dots) and manufacturing areas (blue dots). As shown before, the municipalities
with the highest density grew no faster than the areas with the lowest density.
Municipalities with the highest rate of population growth are Greater Kokstad (Kwazulu Natal), Bitou (Western Cape) and Overstrand (Western Cape). These are predominantly agricultural or manufacturing and wellconnected tourism hubs. Greater Kokstad, for example, is located on the main transport arterial linking KZN
and the Eastern Cape and is the point at which the rail transport link stops. Overstrand is situated on the
‘Cape Whale Coast’ and includes the booming town of Hermanus. Vulamehlo, Impendle and Ndwedwe are
three municipalities with declining populations. Impendle and Ndwendwe are administrative districts and
Vulamehleo has very little formal economic activity.
2
y=1.3823 − 0.0948x, R = 0.0103
Bitou
Overstrand
Greater Kokstad
Population growth (%)
0
5
GRAPH 13: Population density versus population growth
uMhlathuze
Ekurhuleni
Joburg
eThekwini
Emfuleni
−5
Impendle
Ndwedwe
Vulamehlo
0
2
4
6
2
Log (Number of people per km )
Agriculture
Mining
8
Community
Manufacturing
Appendix F presents the regression results when each of the sectors ‘agriculture’, ‘mining’, ‘manufacturing’
and ‘community services’ is treated separately. Statistically significant relationships between density and
growth are only found in the community services sample. It is surprising that no relationship is found in
manufacturing, where it would be most expected.
6.2.3.1. Community services
There were some contradictory relationships among the 23 community services municipalities.
Population growth = 4.3198 - 0.8824 (Log(Population density)), N = 23, R2 = 0.2292
(5)
(1.5785) (0.3531)
GVA growth = 1.1451 + 0.4130 (Log(Employment density)), N = 23, R2 = 0.1378
(6)
(0.2254) (0.3706)
GVA growth
= 3.3036 + 0.3917 (Log(GVA density)),
(2.3708) (1.840)
34
SECTION 6
N = 23, R2 = 0.1774 (7)
A negative relationship was apparent between population density and population growth (see also Graph
14). This is statistically significant at the 5% level. Increasing population density by 1% resulted in a decline
in population of 0.88%. The population decline experienced by the densely populated municipalities
may be attributable to the weak economy of these areas and the pressure on resources. This is possibl­e
bearing in mind that 22 of these 23 municipalities are rural, 18 are former Bantustans and four are
commerci­al farmin­g areas. People may have migrated out of these municipalities in search of better economic opportunitie­s elsewhere.
GRAPH 14: Population density versus population growth
2
2
y = 4.3198 − 0.8824x, R =0.2292
Jozini
Ntabankulu
Greater Giyani
Thulamela
Sol Plaatjie
King Sabata Dalindyebo
Nyandeni
Nquthu Mbizana
Mbhashe
1
Mafikeng
Umzimvubu
Lukhanji
Ulundi
Msinga
EngcoboAganang
Nongoma
0
Population growth (%)
Mutale
Nkandla
−1
Makhuduthamaga
Bushbuckridge
Dr JS Moroka
3.5
4
4.5
5
2
Log (Number of people per km )
5.5
Community services municipality
The relationship between employment density and GVA growth is positive and statistically significant at the
10% level. All else being equal, a 1% increase in employment density in 1996 saw a 0.4% increase in GVA
growth subsequently. The relationship between GVA density and GVA growth is also positive and statistically
significant at the 5% level. A 1% increase in GVA density in 1996 increased GVA growth subsequently by
0.4%. Graph 15 shows the relationship between employment density and GVA growth.
GRAPH 15: Employment density versus economic growth (GVA)
2
y = −3.3036 + 0.3917x, R = 0.1774
4
Greater Giyani
Aganang
GVA growth(%)
2
3
Makhuduthamaga
Mutale
Sol Plaatjie
Dr JS Moroka
1
Nongoma
Msinga
0
Lukhanji
Bushbuckridge
Nyandeni
Mbhashe
Mbizana
Ulundi
Umzimvubu
Nkandla
Jozini
King Sabata Dalindyebo
Thulamela
Mafikeng
1
2
3
2
Log(Number of people employed per km )
4
Community services municipality
SECTION 6
35
6.2.3.2. Summary
This section analysed the relationship between density and growth for municipalities grouped by economic
structure: agriculture, mining, manufacturing, community services. Statistically significant relationships
were only found in the community services category. The results suggest that community services municipalities with higher levels of economic output and employment to begin with experienced higher rates of growth
subsequently. One explanation may be that places with higher economic densities offered larger consumer
markets which attracted more private investment. Another explanation is regional centres of public services
have continued to attract public investment because they are good locations for adminstering such services.
A combination of both explanations is also possible.
6.2.4.Skills
The simple message of the preceding analysis is that the relationship between density and growth is weak or
non-existent, and only appears in particular circumstances. This section examines the relationship between
an area’s skills and its subsequent growth, to see whether this relationship is stronger. Skills is measured by
the proportion of people with matric.
The regressions of skills on growth are presented in the bottow rows of Appendices C, E and F. Across 10
of the 12 groups of municipalities and across all three measures of growth there is a positive relationship
between skills and growth. The higher the share of people with matric in an areas, the stronger its growth
rate. The only exceptions are in the ‘urban’ and ‘metro’ categories. This may have something to do with small
sample sizes in these categories.
The results presented below are for all 237 municipalities. The positive relationship between skills in an
area and its subsequent growth in population, employment and GVA are shown below. They are statistically
significant at the 1% level.
Population growth
= 0.0175 + 0.1489(Skills share), N= 237, R2 = 0.1330
(8)
(0.1967) (0.0248)
Employment growth = 0.0312 + 0.1991(Skills share), N = 237, R2 = 0.0563
(9)
(0.4216) (0.0531)
GVA growth = 0.9778 + 0.1621 (Skills share),N = 237, R2 = 0.1502
(10)
(0.1994) (0.0251)
Holding all else constant, a 1% increase in a municipality’s skills share is associated with a 0.15% increase
in its population growth rate. Graph 16 shows the correlation between skills and population growth. Rural
municipalities have green dots and urban municipalities red dots. Urban areas generally have more skilled
people, although several rural areas are comparable, including Mookgopong (in Limpopo) and Bitou and
Overstrand (both in the Western Cape).
An increase in skills share is also associated with faster employment growth. A 1% increase in the proportion of people with matric increases employment growth by 0.2%. The correlation between skills and
employme­nt growth is shown in Graph 17. The former Bantustans and some commercial farming areas have
relatively low skill levels.
36
SECTION 6
GRAPH 16: Skills share versus economic growth (GVA)
Bitou
2
y = 0.0175+ 0.1489x, R = 0.1330
5
Greater Kokstad
Kungwini
Population growth (%)
0
Mookgopong
Overstrand
uMhlathuze
Midvaal
Johannesburg
Stellenbosch Tshwane
Msunduzi Tlokwe
Emfuleni
−5
Matjhabeng
Richtersveld
Vulamehlo
0
5
10
15
Share of the population with a matric certificate (%)
Rural
20
Urban
15
GRAPH 17: Skills share versus employment growth
2
y = 0.0312 + 0.1991x, R = 0.0563
Employment growth (%)
−5
0
5
10
Mthonjaneni
Mkhambathini
Mtubatuba
uMhlathuze
Bitou
Ekurhuleni
Matlosana
Mookgopong
eThekwini Midvaal
Tshwane
Johannesburg
Stellenbosch
−10
Matjhabeng
Vulamehlo
0
Ndwedwe
Dr JS Moroka
Ngqushwa
Dannhauser
5
10
15
Share of the population with a matric certificate (%)
Bantustan
Secondary
20
Commercial
Metro
Higher skills also means faster GVA growth. Doubling the proportion of people with matric is associated with
a 16% increase in GVA growth (Graph 18). Manufacturing municipalities tend to have higher skills and to have
grown more strongly than other types of area.
Graph 19 shows the relationship between skills and GVA growth analysed separately by sectoral composition
of the area. The association is positive for all types of area and statistically significant for all except mining.
Increasing the share of people with matric has the strongest effect on manufacturing municipalities.
SECTION 6
37
GRAPH 18: Share of the population with matric versus GVA growth
2
Bela−bela
6
y = 0.9778 + 0.1621x, R = 0.1502
Mossel Bay
Knysna Nokeng tsa Taemane
Mookgopong
GVA growth (%)
0
2
4
Tshwane
Johannesburg
Midvaal
Mogale City
Randfontein
Stellenbosch
−2
Emfuleni
−4
Westonaria
Dannhauser
0
Merafong City
Matlosana
5
10
15
Share of the population with a matric certificate (%)
Agriculture
Mining
20
Community
Manufacturing
GRAPH 19: Skills share versus economic growth (GVA)
Community Services
2
y = 0.9470 + 0.1579x, R = 0.1868
2
Manufacturing
Mining
2
y = 0.9069 + 0.1909x, R = 0.3423
2
y = 3.8662 + 0.4875x, R = 0.1204
−5
0
5
10
−5
GVA growth (%)
0
5
10
Agriculture
y = 1.1659 + 0.1589x, R = 0.1386
0
5
10
15
20
0
5
10
Share of the population with a matric certificate (%)
15
20
NOTE
2
Dividing the coefficient by the standard error, we find 0.3653/0.1578=2.31. Since this is greater than 2 we can infer that
density has a significant impact on growth
38
SECTION 6
SECTION
7
CONCLUSION
There is growing interest among governments around the world in the contribution of cities to local and
national economic development. Researchers are also beginning to find evidence that larger and higher
density settlements generate higher levels of productivity and growth. However, this relationship is best
interpreted as a tendency rather than a universal law because the research suggests that there are sizeable differences in the nature and magnitude of this effect, depending on the context and the way in which
density and growth are measured. It seems premature to argue that concentrated economic activity through
urbanisation necessarily increases national economic growth.
This paper examines whether the density of population and economic activity influences the rate of local
economic development in South Africa. Municipalities are the basic units of analysis and the time frame is
1996-2010. Contrary to expectations, no statistically significant relationship is found between density and
development when the analysis covers all 237 local municipalities. There is no evidence that municipalities
with higher densities of population or economic activity in 1996 grew any faster than municipalities with
lower densities over the subsequent 15 year period.
Searching hard for a relationship we examined particular types of municipality on their own to isolate other
factors. This resulted in some evidence of a relationship between density and growth emerging for a few
select groups. Statistically significant relationships between density and economic growth are apparent for
the commercial farming areas and for municipalities dominated by community services sectors. Once again,
this is not what one would have expected. Previous research suggests that agglomeration effects should be
most apparent in (large) urban areas and in areas with sizeable industrial sectors.
By way of a comparison, the paper also examines the influence of human skills on local economic growth.
The relationship between skills and growth is found to be more robust than density. There are positive,
statistically significant relationships across most groups of municipalities. When all 237 municipalities are
considered, a 1% increase in the proportion of the population with matric is associated with a 0.2% increase
in employment and GVA growth. It seems clear that skills is strongly linked with economic growth.
An important message from the analysis is that spatial proximity does not inevitably or automatically contribute to local economic growth. There is no necessary connection between the density of activity within an
area and its rate of economic growth.
There are several possible reasons why the relationship between density and growth is weak or non-existent
in South Africa. First, the legacy of the colonial and Apartheid spatial policies, including forced removals
and restrictions on mobility, may have disrupted conditions by creating artificially high population densities
39
in the former Bantustans. Many of these areas are physically isolated, agriculturally infertile and economically unproductive. Second, the impact of racial segregation and residential restrictions within cities and
towns may have resulted in more spatially fragmented and inefficient urban areas, thereby undermining
their economic performance. Third, the equitable sharing of government tax revenues and fiscal transfers to
compensate poorer districts may mask the impact of density and lift the performance of weak local economies. Fourth, there may be reliability problems with the dataset, bearing in mind its uncertain origins. The
requirement that the local statistics are adjusted to add up to the national totals may also introduce some
kind of systematic distortion. Against this, the discovery of a significant relationship between skills and
growth provides some reassurance that the dataset should not be discounted.
Further research is required to evaluate the relevance of these factors and to assess whether there is an
underlying relationship between density and growth once these influences have been discounted. The
research could also be extended through multivariate regression. This would enable the various influences
to be incorporated into the analysis simultaneously, thereby identifying their relative importance and the
ways in which they interact. An additional refinement would be to incorporate the detailed industrial composition of each area without the simplification of creating four municipal categories of manufacturing, mining,
agriculture and community services.
40
SECTION 7
BIBLIOGRAPHY
Abel, J. R., Dey, I., and Gabe, T. M. (2011). Productivity and the Density of Human Capital. Journal of Regional Science, 1-25.
Baldwin, A. (1975). Mass Removals and Separate Development. Journal of Southern African Studies, 215-227.
Banjeree, A., Galiani, S., Levinsohn, J., and Woolard, I. (2007). Why has unemployment risen in the New South Africa. he Economics of Transition, The European Bank for Reconstruction and Development, 715-740.
Bigsten, A., Gebreeyesus, M., Siba, E., and Soderbom, M. (2011). The Effects of Agglomeration and Competition on Prices and
Productivty : Evidence for Ethiopia’s manufacturing Sector. University of Gothenburg.
Carlin, W. (2008). University College London. Retrieved February 15, 2012, from http://www.ucl.ac.uk/~uctpa36/C41_note_
growth_accounting.pdf
Chen, Y. (1996). Impact of Regional Factors on Productivity in China. Journal of Regional Science, 417-36.
Ciccone, A. (2002). Agglomeration effects in Europe. European Economic Review, 213-227.
Ciccone, A., and Hall, R. (1996). Productivity and the density of economic activity. American Economic Review, 54-70.
Cingano, F., and Schivardi, F. (2003). Identifying the Sources of Local Productivity Growth. Rome: Bank of Italy.
Cota, J. (2001). Specialisation, Agglomeration and Urban Manufacturing in the Northern Border Cities of Mexico. Journal of Borderlands Studies, 71-97.
Deichmann, U., Gill, I., and Goh, C. (2011). ‘Texture and tractability: The framework for spatial policy analysis in the World Development Report 2009’. Journal of Regions, Economy and Society, 163-174.
Duranton, G., and Puga, D. (2001). Nursery cities: Urban diversity, process innovation,. American Economic Review, 1454–1477.
Duranton, G. and H. G. Overman (2002). Testing for localisation using micro-geographic
data,. London School of Economics Working Paper.
Fafchamps, M. (2004). Manufacturing Growth and Agglomeration Effects. Oxford: University of Oxford CSAE Working Paper
2004-33.
Glaeser, E. (2007). The Economic Approach to Cities. NBER Working Paper 13696.
Glaeser, E. (2011) Triumph of the City, London: Macmillan.
Glaeser, E., Kallal, H., Scheinkman, H., and Shleifer, A. (1992). Growth in Cities. Journal of Political Economy, 1126-1152.
Glaeser, E. L., and Kahn, M. E. (2003). Sprawl and Urban Growth. NBER Working paper series.
Glaeser, E., and Gottlieb, J. (2009). The Wealth of Cities: Agglomeration Economies and Spatial Equilibrium in the United States.
Journal of Economic Literature, American Economic Association, 983-1028.
Global Insight. (2011). IHS Global Insight Regional Explorer Encyclopedia. IHS Global Insight.
Graham, D. (2007). Agglomeration Economies and Transport Investment. OECD European Conference of Ministers of Transport,
17 September.
Graham, D. (2009). Identifying urbanisation and localisation externalities in manufacturing and service industries. Papers in
Regional Science, 63-84.
Gujarati, D. (2003). Basic Econometrics. New York: McGraw-Hill/Irwin
Henderson, V. (1986). Efficiency of Resource Usage and City Size. Journal of Urban Economics, 47-70.
Henderson, V. (1988). Urban Development: Theory, Fact, and Illusion. Oxford: Oxford University Press.
Henderson, V., and Kuncoro, A. (1996). Industrial Centralization in Indonesia. World Bank Economic Review, 513-40.
Henderson, V. (2002). Urbanisation in Developing Countries. The World Bank Observer, 89-112.
Henderson, V., Lee, T., and Lee, Y. (2001). Scale Economies in Korea. Journal of Urban Economics, 479-504.
Jacobs, J. (1984). Cities and the Wealth of Nations: Principles of Economic Life. New York: Random House.
Krugman, P. (1991). Increasing Returns and Economic Geography. Journal of Political Economy, 483-499.
41
Krugman, P. (1998). What’s New About New Economic Geography? Oxford Review of Economic Policy, 7-17.
Lee, Y.-J., and Zang, H. (1998). Urbanisation and Regional Productivity in Korea. Urban Studies, 2085-99.
Makgetla, N. (2010). Synthesis Paper: South Africa. Unpublished Paper.
Mitra, A. (2000). Total Factor Productivity Growth and Urbanisation Economies: A Case of Indian Industries. Review of Urban and
Regional Development Studies, 97-108.
Munoz, M. M., Gill, I., and Goh, C.-c. (2009). Integration through institutions, infrastructure and interventions. Rural 21, 22-25.
Overman, H., and Venables, A. (2005). Cities in the Developing World. London: Centre for Economic Performance Discussion
Paper.
Overman, H. and Venables, A. (2010) ‘Evolving city systems’, chapter in Beall, J., Guha-Khasnobis, B. and Kanbur, R. (eds) (2010)
Urbanisation and Development: Multidisciplinary Perspectives, Oxford: Oxford University Press
Quigley, J. (2008). Urbanization, Agglomeration and Economic Development, Working Paper 19. Washington: The World Bank.
Roberts, M., and Goh, C.-c. (2010). Density, distance and division: the case of Chongqing municipality, China. Cambridge Journal
of Regions, Economy and Society, 189-204.
Rosenthal, S., and Strange, W. (2003). Geography, Industrial Organization and Agglomeration. The Review of Economics and
Statistics, 377-393.
Rosenthal, S., and Strange, W. (2004). Evidence on the Nature and Sources of Agglomerarion Economies. Handbook of Regional
and Urban Economics, 2119-2171.
Schwabe, C. (2004). Fact Sheet: Poverty in South Africa. Human Sciences Research Council (HSRC).
Soest, D., Gerking, S., and Oort, F. (2002). Knowledge Externalities, Agglomeration Economies, and Employment Growth in Dutch
Cities. Tilburg University: Centre for Economic Research Discussion Paper 2002-41.
Wittenberg, M. (2010) Econometrics through applications: A practical handbook. Cape Town: UCT.
World Bank (2008) World Development Report 2009: Reshaping Economic Geography. Washington DC: The World Bank.
42
BIBLIOGRAPHY
APPENDICES
APPENDIX A
Population
Global Insight determines national population projections by five primary factors:
•
Size of population in the base year, Pt
•
Number of deaths occurring between the base and projected years, Dt
•
Number of births occurring between the base and projected years, Bt
•
Immigrants arriving in the country between the base and projected years, It
•
Emigrants leaving the country during the base and projected years, Et
These variables contribute to the projected population, Pt +1, within the following demographic balancing
identity
Pt + 1 = Pt + Bt – Dt + It – Et
Census data and factor-based backward extrapolation is then used to arrive at a 1970 base population figure. This is used to estimate the national base population figure and the base population estimates for each
province. Municipal populations are then estimated by adjusting provincial factors and assumptions based
on underlying provincial evidence.
Employment
Global Insight uses a Labour Model built on two pillars. One estimates formal and informal employment (i.e.
the demand side), while the other estimates unemployment and economic activity (i.e. the supply side). They
do this because data from employers is believed to be more reliable than data from home-based surveys.
Unemployment is measured at the place of residence, while employment is measured at the place of work.
The estimates for each area are balanced and checked against the Quarterly Labour Force Survey and the
General Household Survey. GI also obtains regional employment data from relevant industry associations
and interpolates for the missing figures on the basis of the relevant sector’s output growth in each region.
Employment numbers are then estimated so that the following labour market identities balance: EAP = U + E
Where:
•
EAP = Total Economically Active Population
•
U = Number of people unemployed
•
E = Number of people employed (formal + informal sector)
Gross Value Added (GVA)
GI obtains initial estimates of GVA growth rates from five sources: mining, construction, electricity, retail
trade and regional service council levies. These growth rates are applied to preliminary estimates of GVA
benchmarked on national level Reserve Bank estimates of value added by sector to arrive at preliminary
43
estimates of GVA for each year from 1997 to 2005. These estimates are then benchmarked and adjusted
to national level estimates of sectoral GVA (unpublished detailed series obtained from StatsSA as well as
Reserve Bank published series) to arrive at final regional estimates.
APPENDIX B
Categorisation of municipalities by their economic base
The overriding principle is that municipalities are classified according to their dominant tradable sector.
We start off by identifying municipalities with more than 10% of their jobs in mining. In 1996, there were 34
municipalities with this characteristic (referred to below as group 1). To establish whether these are classified
as mining or another sector, we compare their share of jobs in mining with their share of jobs in agriculture
and in manufacturing. If the mining share exceeds the other sectors, these areas are classified as mining.
1.
Of the 34 municipalities in group 1, eight have both
• agricultural share of employment less than 10% and
• manufacturing share of employment less than 10%.
Therefore these municipalities are classified as mining
Municipality
Sector
Agriculture
Mining
Manufacturing
Community
Westonaria
Mining
1%
75%
3%
4%
Merafong City
Mining
2%
68%
4%
6%
Matjhabeng
Mining
3%
61%
3%
9%
City of Matlosana (Klerksdorp)
Mining
4%
55%
4%
10%
Kgetlengrivier
Mining
6%
47%
6%
10%
Rustenburg
Mining
7%
46%
6%
10%
Moses Kotane
Mining
7%
23%
8%
22%
Fetakgomo
Mining
1%
17%
2%
43%
Classified
8
0
8
0
0
2.
Of the remaining 26 municipalities from group 1, five have
• Manufacturing share of employment greater than 10% and
• Agricultural share of employment less than 10%.
These municipalities are therefore classified as manufacturing or mining depending on
which sector was larger. Four were classified as mining and one manufacturing.
Municipality
Sector
Agriculture
Mining
Manufacturing
Community
Randfontein
Mining
5%
43%
12%
10%
Govan Mbeki (Highveld)
Mining
5%
36%
20%
9%
Dannhauser
Mining
6%
35%
16%
13%
Emalahleni(Mpumalanga)
Mining
4%
23%
19%
11%
Manufacturing
8%
14%
17%
20%
5
0
4
1
0
NokengtsaTaemane
Classified
44
APPENDICES
3.
Of the remaining 21 municipalities from group 1, 16 municipalities have
• Agricultural share greater than 10% and
• Manufacturing share of employment less than 10%
These municipalities are therefore classified as mining or agriculture depending on which
sector was larger. 10 municipalities are classified as mining and 6 as agriculture.
Municipality
Sector
Agriculture
Mining
Manufacturing
Community
Thabazimbi
Mining
13%
63%
3%
4%
Masilonyana
Mining
27%
40%
2%
9%
Kgatelopele (Dan-Lime)
Mining
12%
36%
6%
16%
Tsantsabane
Mining
12%
36%
6%
16%
Gamagara
Mining
15%
32%
5%
15%
Richtersveld
Mining
13%
30%
4%
18%
Kamiesberg
Mining
13%
30%
4%
18%
NamaKhoi
Mining
13%
30%
4%
18%
Ba-Phalaborwa
Mining
19%
21%
4%
17%
Ga-Segonyana
Agriculture
22%
18%
3%
21%
Emadlangeni (Utrecht)
Agriculture
38%
16%
4%
18%
Mining
12%
15%
6%
27%
Khai-Ma
Agriculture
50%
13%
2%
12%
Letsemeng
Agriculture
41%
13%
3%
11%
Delmas
Agriculture
28%
11%
9%
10%
Abaqulusi
Agriculture
22%
10%
8%
19%
Classified
16
6
10
0
0
Greater Tubatse
4.
The five remaining municipalities in group 1 had both
• Agricultural share of employment greater than 10% and
• Manufacturing share of employment greater than 10%
These municipalities are therefore classified as manufacturing, mining or agriculture
depending on which sector was the largest. 2 municipalities are classified as agriculture and
3 as mining.
Municipality
Sector
Agriculture
Mining
Manufacturing
Community
Lekwa (Standerton)
Mining
23%
27%
10%
10%
Dikgatlong
Mining
15%
25%
14%
13%
Steve Tshwete (Middelburg)
Mining
11%
23%
15%
12%
Madibeng
Agriculture
19%
10%
15%
17%
Umjindi
Agriculture
45%
10%
13%
8%
5
2
3
0
0
Classified
The 203 remaining municipalities all have less than 10% of employment in mining.
APPENDICES
45
5.
Group 2 consists of these municipalities with more than 10% of jobs in agriculture. Of these
157 municipalities, 100 have less than 10% of their jobs in manufacturing. So these municipalities are classified as agriculture.
Municipality
46
APPENDICES
Sector
Agriculture
Mining
Manufacturing
Community
Witzenberg
Agriculture
65%
0%
8%
9%
Siyancuma
Agriculture
59%
1%
6%
13%
Sunday’s River Valley
Agriculture
58%
0%
4%
12%
Tokologo
Agriculture
57%
6%
2%
8%
Theewaterskloof
Agriculture
56%
0%
7%
11%
Cederberg
Agriculture
55%
0%
6%
10%
Ventersdorp
Agriculture
54%
2%
5%
10%
Kannaland
Agriculture
54%
0%
9%
13%
Kou-Kamma
Agriculture
52%
0%
7%
9%
Renosterberg
Agriculture
52%
0%
1%
16%
Karoo Hoogland (Frasuwil)
Agriculture
52%
0%
2%
12%
Nxuba
Agriculture
52%
0%
1%
22%
Baviaans
Agriculture
51%
0%
3%
19%
Prince Albert
Agriculture
51%
0%
3%
15%
Setsoto
Agriculture
50%
0%
5%
11%
Tswelopele
Agriculture
49%
0%
4%
11%
Maquassi Hills
Agriculture
48%
4%
4%
11%
Tsolwana
Agriculture
48%
0%
4%
23%
Siyathemba
Agriculture
48%
0%
5%
16%
Mamusa (Schweizer-Reneke)
Agriculture
47%
4%
4%
10%
Impendle
Agriculture
47%
0%
6%
17%
Matzikama
Agriculture
46%
4%
6%
10%
Richmond
Agriculture
46%
0%
10%
13%
Nala
Agriculture
46%
1%
6%
12%
Blue Crane Route
Agriculture
46%
0%
6%
19%
!Kai! Garib
Agriculture
45%
2%
4%
15%
Tswaing
Agriculture
43%
7%
4%
13%
Naledi
Agriculture
43%
1%
5%
12%
Naledi (Free State)
Agriculture
43%
0%
3%
16%
Modimolle
Agriculture
43%
1%
7%
14%
!Kheis
Agriculture
43%
2%
4%
16%
Kagisano
Agriculture
42%
0%
4%
24%
Molopo
Agriculture
41%
0%
3%
28%
Ikwezi
Agriculture
41%
0%
3%
17%
Mantsopa
Agriculture
41%
0%
5%
15%
Ubuntu
Agriculture
40%
0%
3%
18%
Musina
Agriculture
40%
4%
5%
13%
Ndlambe
Agriculture
39%
1%
7%
13%
Kouga
Agriculture
38%
0%
10%
12%
Nketoana
Agriculture
38%
0%
4%
13%
Mohokare
Agriculture
38%
0%
3%
17%
Mafube
Agriculture
38%
0%
5%
15%
Gariep (Eastern Cape)
Agriculture
38%
0%
8%
19%
Phokwane
Agriculture
38%
1%
5%
15%
Dipaleseng
Agriculture
38%
4%
7%
12%
Kareeberg
Agriculture
37%
0%
4%
20%
KharaHais
Agriculture
37%
3%
5%
17%
Mier
Agriculture
37%
3%
5%
17%
Hantam
Agriculture
37%
0%
6%
17%
Hessequa (Langeberg)
Agriculture
37%
0%
7%
14%
Greater Letaba
Agriculture
36%
0%
7%
16%
eDumbe
Agriculture
36%
5%
9%
13%
Mookgopong
Agriculture
35%
1%
8%
13%
Kopanong
Agriculture
35%
0%
3%
20%
Elias Motsoaledi
Agriculture
34%
3%
5%
20%
Cape Agulhas
Agriculture
33%
0%
9%
15%
Dihlabeng
Agriculture
33%
0%
6%
15%
Greater Kokstad
Agriculture
33%
0%
6%
14%
PixleyKaSeme
Agriculture
32%
2%
5%
13%
Magareng
Agriculture
32%
2%
6%
19%
Umsobomvu
Agriculture
32%
0%
2%
21%
Albert Luthuli
Agriculture
31%
7%
5%
23%
Thembelihle
Agriculture
31%
0%
5%
22%
Greater Tzaneen
Agriculture
31%
1%
8%
18%
Phumelela
Agriculture
31%
0%
8%
13%
Msukaligwa
Agriculture
31%
8%
7%
14%
Greater Marble Hall
Agriculture
31%
1%
5%
23%
Ngwathe
Agriculture
30%
0%
9%
16%
Makhado
Agriculture
29%
1%
7%
20%
InxubaYethemba
Agriculture
28%
2%
5%
24%
RamotshereMoiloa (Zeerust)
Agriculture
28%
1%
5%
28%
Bela-Bela
Agriculture
27%
0%
7%
13%
Maruleng
Agriculture
26%
0%
6%
20%
Lekwa-Teemane
Agriculture
26%
2%
9%
16%
Senqu
Agriculture
26%
0%
3%
32%
Moqhaka
Agriculture
25%
4%
9%
19%
Sakhisizwe
Agriculture
24%
0%
4%
30%
Beaufort West
Agriculture
23%
0%
5%
22%
Makana
Agriculture
23%
0%
7%
33%
APPENDICES
47
Lephalale
Agriculture
23%
9%
4%
11%
Elundini
Agriculture
22%
0%
5%
30%
Emalahleni
Agriculture
21%
0%
5%
38%
Maletswai
Agriculture
20%
0%
8%
25%
Moshaweng
Agriculture
20%
5%
3%
41%
Camdeboo
Agriculture
19%
0%
8%
23%
Molemole
Agriculture
18%
1%
8%
20%
The Big Five False Bay
Agriculture
18%
0%
7%
33%
Emthanjeni
Agriculture
17%
0%
5%
25%
Umzimkhulu (Umzimkulu)
Agriculture
17%
1%
9%
38%
Nkonkobe
Agriculture
16%
0%
4%
47%
Ratlou (Setla-Kgobi)
Agriculture
16%
0%
6%
32%
Mogalakwena
Agriculture
15%
1%
2%
11%
Blouberg
Agriculture
15%
1%
5%
37%
Matatiele
Agriculture
14%
0%
4%
31%
Polokwane
Agriculture
14%
0%
8%
21%
Mhlontlo
Agriculture
12%
0%
6%
43%
IntsikaYethu
Agriculture
12%
0%
3%
42%
Lepelle-Nkumpi
Agriculture
12%
1%
3%
19%
Greater Taung
Agriculture
11%
2%
6%
39%
Umhlabuyalingana
Agriculture
10%
1%
5%
39%
100
100
0
0
0
Classified
6.
The remaining 57 municipalities in group 2 have
• Agricultural share of employment greater than 10%, and
• Manufacturing share of employment greater than 10%.
These municipalities are therefore classified as agriculture or manufacturing depending on
which sector was larger. Stellenbosch is categorised as agriculture.
Sector
Agriculture
Mining
Manufacturing
Community
Bergrivier
Agriculture
54%
0%
11%
10%
Breede River/Winelands
Agriculture
51%
0%
13%
11%
Swellendam
Agriculture
47%
0%
10%
11%
uMshwathi
Agriculture
44%
1%
22%
11%
Mooi Mpofana
Agriculture
44%
0%
17%
12%
Laingsburg
Agriculture
44%
1%
10%
13%
KwaSani
Agriculture
43%
0%
11%
9%
Umvoti
Agriculture
42%
0%
12%
14%
Breede Valley
Agriculture
42%
0%
13%
16%
Mthonjaneni
Agriculture
38%
0%
15%
19%
KwaDukuza
Agriculture
38%
0%
22%
9%
Mkhondo
Agriculture
36%
5%
14%
12%
Ubuhlebezwe
Agriculture
35%
0%
10%
19%
Inkwanca
Agriculture
35%
0%
13%
21%
Municipality
48
APPENDICES
Ingwe
Agriculture
33%
0%
10%
21%
Nkomazi
Agriculture
31%
6%
11%
18%
Ndwedwe
Agriculture
30%
0%
16%
22%
Hlabisa
Agriculture
29%
0%
11%
24%
ThabaChweu
Agriculture
29%
6%
11%
13%
Maphumulo
Agriculture
29%
0%
13%
30%
Drakenstein
Agriculture
29%
0%
24%
15%
Mtubatuba
Agriculture
27%
1%
12%
23%
Swartland
Manufacturing
27%
0%
31%
11%
Oudtshoorn
Agriculture
26%
0%
11%
23%
Ditsobotla (Lichtenburg)
Agriculture
25%
4%
12%
18%
Manufacturing
24%
0%
30%
16%
Emakhazeni (Highlands)
Agriculture
24%
9%
12%
12%
Okhahlamba
Agriculture
24%
0%
11%
24%
Mkhambathini
Agriculture
23%
0%
20%
18%
Overstrand
Agriculture
22%
0%
11%
14%
Umzumbe (Khiphinkunzi)
Agriculture
21%
0%
17%
18%
uMuziwabantu
Agriculture
21%
0%
16%
21%
Ezingoleni (Izingolweni)
Agriculture
20%
1%
15%
20%
Stellenbosch
Agriculture
20%
0%
20%
22%
Umdoni
Manufacturing
20%
0%
22%
18%
uMngeni
Agriculture
19%
0%
17%
18%
Mbombela
Agriculture
19%
1%
13%
16%
Vulamehlo
Manufacturing
18%
0%
21%
21%
uPhongolo
Agriculture
18%
5%
12%
21%
Hibiscus Coast
Agriculture
18%
1%
12%
16%
Saldanha Bay
Manufacturing
17%
1%
27%
16%
uMlalazi
Manufacturing
17%
1%
35%
19%
Ntambanana
Manufacturing
16%
4%
18%
16%
Agriculture
15%
3%
11%
25%
Great Kei
Manufacturing
14%
0%
21%
18%
Endumeni
Agriculture
14%
4%
11%
24%
uMhlathuze
Manufacturing
14%
4%
20%
18%
George
Manufacturing
13%
0%
17%
18%
Lesedi
Manufacturing
12%
1%
21%
19%
Umtshezi
Manufacturing
12%
0%
25%
19%
Indaka
Manufacturing
11%
2%
19%
22%
Mossel Bay
Manufacturing
11%
3%
21%
17%
Amahlati
Manufacturing
11%
0%
22%
34%
Kungwini
Manufacturing
11%
2%
15%
15%
Knysna
Manufacturing
11%
0%
16%
14%
Bitou (Plettenberg Bay)
Manufacturing
11%
0%
16%
14%
Port St Johns
Manufacturing
10%
0%
13%
40%
57
39
0
18
0
Mandeni
(Endondakusuka)
Tlokwe (Potchefstroom)
Classified
APPENDICES
49
7.
Group 3 consists of municipalities with more than 10% of their jobs in manufacturing. All
of these 23 areas have less than 10% of their jobs in agriculture, so they are classified as
manufacturing.
Municipality
Sector
Agriculture
Mining
Manufacturing
Community
Newcastle
Manufacturing
3%
1%
36%
20%
Emfuleni
Manufacturing
3%
0%
35%
15%
Emnambithi-Ladysmith
Manufacturing
4%
0%
35%
18%
Moretele
Manufacturing
1%
0%
32%
22%
N Mandela
Manufacturing
2%
0%
31%
21%
Metsimaholo
Manufacturing
6%
1%
30%
14%
eThekwini
Manufacturing
2%
0%
29%
18%
Imbabazane
Manufacturing
9%
0%
28%
19%
Ekurhuleni
Manufacturing
1%
3%
27%
14%
Buffalo City
Manufacturing
4%
0%
27%
27%
City of Cape Town
Manufacturing
2%
0%
26%
20%
Ngqushwa
Manufacturing
5%
0%
24%
40%
Midvaal
Manufacturing
4%
0%
21%
16%
Thembisile
Manufacturing
5%
2%
21%
28%
Msunduzi
Manufacturing
3%
0%
21%
24%
Mogale City
Manufacturing
8%
4%
20%
18%
Mbonambi
Manufacturing
10%
5%
19%
15%
City of Johannesburg
Manufacturing
1%
1%
19%
17%
Maluti a Phofung
Manufacturing
9%
0%
18%
27%
Mnquma
Manufacturing
4%
0%
15%
40%
City of Tshwane
Manufacturing
1%
0%
15%
25%
Ngquza Hill
Manufacturing
9%
0%
14%
38%
Mangaung
Manufacturing
5%
0%
13%
28%
Classified
23
0
0
23
0
8.
The 23 remaining municipalities have less than 10% of their jobs in agriculture, mining and
manufacturing, and more than 10% of their jobs in community services, so they are classified
in the latter category.
Sector
Agriculture
Mining
Manufacturing
Community
Bushbuckridge
Community
9%
0%
9%
39%
Sol Plaatjie
Community
5%
3%
9%
30%
Dr JS Moroka
Community
7%
0%
9%
40%
Thulamela
Community
6%
1%
9%
38%
Lukhanji
Community
7%
0%
8%
33%
Aganang
Community
3%
0%
8%
39%
Mafikeng
Community
7%
0%
7%
36%
King
SabataDalindyebo
Community
3%
0%
6%
39%
Municipality
50
APPENDICES
Masilonyana
Masilonyana
Dannhauser
Dannhauser
Dikgatlong
Lekwa
Kamiesberg
6%
43%
9%
1%
6%
40%
5%
0%
5%
41%
Community
8%
0%
5%
57%
Ulundi
Community
4%
6%
4%
49%
Greater Giyani
Community
9%
1%
4%
39%
Mutale
Community
8%
4%
4%
47%
Nquthu
Community
9%
1%
4%
48%
Mbizana
Community
4%
1%
3%
39%
Msinga
Community
6%
1%
3%
47%
Makhuduthamaga
Community
5%
7%
3%
46%
Ntabankulu
Community
6%
1%
3%
46%
Nongoma
Community
6%
3%
3%
46%
Umzimvubu
Community
7%
0%
3%
42%
Mbhashe
Community
6%
1%
2%
46%
Classified
23
0
0
0
23
Fetakgomo
Fetakgomo
Greater Tubatse
Greater Tubatse
Ba-Phalaborwa
Moses Kotane
Steve Tshwete
0%
Ba-Phalaborwa
Moses Kotane
Steve Tshwete
Emalahleni (Mpumalanga)
alahleni (Mpumalanga)
Dikgatlong
Lekwa
Kamiesberg
Nkandla
Nama Khoi
Community
Nama Khoi
Community
Engcobo
Richtersveld
Jozini
Richtersveld
7%
Gamagara
Kgatelopele
Kgatelopele
Community
Gamagara
Tsantsabane
Tsantsabane
Govan Mbeki
Randfontein
Randfontein
Govan Mbeki
Rustenburg
Kgetlengrivier
Matlosana
Rustenburg
Kgetlengrivier
Matlosana
Matjhabeng
Thabazimbi
Merafong City
0
Matjhabeng
10
0
Thabazimbi
0
Merafong City
30 0
Witzenberg
Witzenberg
Tokologo
Tokologo
Ventersdorp
Ventersdorp
Kou-Kamma
Kou-Kamma
Renosterberg
Renosterberg
Prince Albert
Prince Albert
Tswelopele
Tswelopele
Maquassi Hills
Maquassi Hills
Swellendam Swellendam
Richmond Richmond
!Kai! Garib !Kai! Garib
Laingsburg Laingsburg
Modimolle Modimolle
Naledi
Naledi
Breede Valley
Breede Valley
Mantsopa Mantsopa
Ubuntu
Ubuntu
Gariep (Eastern
Cape)
Gariep
(Eastern Cape)
Mafube
Mafube
Kwa Sani Kwa Sani
Hantam
Hantam
Khara Hais Khara Hais
Greater Letaba
Greater Letaba
Kopanong Kopanong
Elias Motsoaledi
Elias Motsoaledi
Greater Kokstad
Greater Kokstad
UmsobomvuUmsobomvu
Greater Tzaneen
Greater Tzaneen
Msukaligwa Msukaligwa
Ngwathe
Ngwathe
Maphumulo
Maphumulo
Drakenstein
Drakenstein
Ramotshere Moiloa
Ramotshere Moiloa
Senqu
Senqu
Oudtshoorn
Oudtshoorn
Sakhisizwe
Sakhisizwe
Makana
Makana
Beaufort West
Beaufort West
Ga-Segonyana
Ga-Segonyana
Umzumbe (Khiphinkunzi)
Umzumbe (Khiphinkunzi)
Ezingoleni (Izingolweni)
Ezingoleni (Izingolweni) Camdeboo
Camdeboo Madibeng
Madibeng
The
Big Five False Bay/
The Big Five False Bay/ Emthanjeni
Emthanjeni
Blouberg
Blouberg Matatiele
MatatieleIntsika Yethu
Greater Taung
Intsika Yethu
Greater Taung
10
Westonaria
20
Westonaria
20
Nyandeni
Graphs B1 to B4 show the distribution of municipalities by sector.
70
60
GRAPH
B1: Share of employment in agriculture for ‘agriculture’ municipalities. N=146
50
70
40
60
30
50
20
4010
GRAPH B2: Share of employment in mining for mining municipalities. N=25
80
70
60
80
50
70
40
60
30
50
20
40
10
30
APPENDICES
51
010
5
0
52
APPENDICES
Indaka
Mangaung
Knysna
Bitou (Plettenberg
Bay)
Mnquma
Knysna
Kungwini
Mnquma
City of Tshwane
Kungwini
Ngquza Hill
City of Tshwane
Port St Johns
Ngquza Hill
Mangaung
Port St Johns
George
Nokeng
tsa Taemane
Bitou (Plettenberg Bay) George
Ntambanana
Maluti a Phofung
Nokeng tsa Taemane
Ntambanana
Maluti a PhofungMbonambi
Mbonambi
uMhlathuze
Johannesburg
Johannesburg
Indaka
GRAPH B4: Share of employment in manufacturing for :manufacturing’ municipalities. N=43
Jozini
Engcobo
Umzimvubu
Nyandeni
Mbizana
Dr JS Moroka
Dr JS Moroka
Jozini
Engcobo
Umzimvubu
Nyandeni
Makhuduthamaga
Makhuduthamaga
Nongoma
Ntabankulu
Ntabankulu
Nongoma
Mbhashe
Mutale
Mutale
Mbhashe
Msinga
Nquthu
Ulundi
Msinga
Nquthu
Ulundi
Aganang
Sol Plaatjie
Lukhanji
Mafikeng
Sol Plaatjie
Lukhanji
Mafikeng
Thulamela
Bushbuckridge
Thulamela
Bushbuckridge
Greater Giyani
Aganang
Mbizana
King Sabata Dalindyebo
King Sabata Dalindyebo
Greater Giyani
40
Mossel Bay
Mogale City
Mogale City
uMhlathuze
Mossel BayThembisile
Umtshezi Umtshezi
Ngqushwa
Ngqushwa
Amahlati
Amahlati
Umdoni
Umdoni
Great Kei
Great Kei
Midvaal
Midvaal
Lesedi
Lesedi
Vulamehlo
Vulamehlo
Msunduzi
Thembisile Msunduzi
Cape Town
Cape Town
EkurhuleniEkurhuleni
Saldanha Saldanha
Bay
Bay
Imbabazane
Imbabazane
Buffalo City
Buffalo City
eThekwinieThekwini
Mandeni Mandeni
SwartlandSwartland
Metsimaholo
Metsimaholo
Moretele Moretele
N MandelaN Mandela
0
Nkandla
10
Nkandla
0
uMlalazi uMlalazi
Newcastle
Newcastle
Emfuleni
Emfuleni
Emnambithi-Ladysmith
Emnambithi-Ladysmith
GRAPH B3: Share of employment in comunity services for “community services” municipalities. N=23
60
50
40
60
50
30
40
20
30
10
20
35
30
40
2535
2030
1525
1020
515
APPENDIX C
Dependent
variable
log (popdensity)
popgrowth
National
Rural
Urban
N=237
N=210
N=27
emptgrowth
Gvagrowth popgrowth
emptgrowth
gvagrowth
Popgrowth
emptgrowth gvagrowth
-0.0948
0.0118
0.0024
-0.1701
-0.0856
-0.0892
-0.0593
-0.0404
0.2334
(0.0608)
(0.1255)
(0.0626)
(0.0722)
(0.1519)
(0.0725)
(0.2118)
(0.3559)
(0.2699)
1.3823
1.3994
2.1155
1.5461
1.6050
2.3276
1.6896
2.3720
1.3960
(0.2214)
(0.4573)
(0.2279)
(0.2413)
(0.5079)
(0.2426)
(1.5029)
(1.9325)
(1.4659)
R
0.0103
0.0000
0.0000
0.0260
0.0015
0.0072
0.0031
0.0005
0.0290
log(emptdensity)
0.0799
0.1548
0.1039
0.7329
0.1031
0.0347
-0.0837
-0.1073
0.1776
(0.0621)
(0.1276)
(0.0634)
(0.0849)
(0.1766)
(0.0846)
(0.2078)
(0.3492)
(0.2669)
0.9604
-0.0298
1.9806
0.9554
1.2393
2.0215
1.7104
2.5879
1.9239
(0.1280)
(1.4889)
0.1309
(0.1359)
(0.2829)
(0.1356)
(0.8661)
(1.4551)
(1.1122)
R
0.0070
0.0042
0.0113
0.0036
0.0016
0.0008
0.0064
0.0038
0.0174
log(gvadensity)
0.0714
0.1150
0.0881
0.0536
0.0305
0.0136
0.0262
0.0497
0.2454
(0.0562)
(0.1156)
(0.0574)
(0.0774)
(0.1610)
(0.0771)
(0.2063)
(0.3461)
(0.2619)
Constant
2
Constant
2
Constant
R2
Share_matric
Constant
R2
0.1589
-0.0298
0.9993
0.3674
0.9676
1.8890
0.9609
1.3731
-1.2266
(0.7236)
(1.4888)
(0.7400)
(0.9637)
(2.0054)
(0.9606)
(3.2579)
(5.4659)
(4.1364)
0.0068
0.0042
0.0099
0.0023
0.0002
0.0001
0.0006
0.0008
0.0339
0.1489***
0.1991***
0.1621***
0.2032***
0.2375***
0.1993***
0.0959
0.1862
0.1576
(0.0248)
(0.0531)
(0.0251)
(0.0315)
(0.0699)
(0.0315)
(0.0854)
(0.1421)
(0.1086)
0.0175
0.0312
0.9778
-0.2575
-0.1601
0.7940
0.1536
-0.2137
0.6299
(0.1967)
(0.4216)
(0.1994)
(0.2216)
(0.4909)
(0.2213)
(1.1110)
(1.8482)
(1.4120)
0.1330
0.0563
0.1502
0.1661
0.0525
0.1611
0.0479
0.0642
0.0778
Notes: Bivariate regression takes the form y= b1x + u. Popgrowth, emptgrowth and GVAgrowth are calculated using the equation growth=
(ln (Xi10 /Xi96)/15)*100 Log(popdensity) = Log(Number of people per km2) Log(emptdensity)= Log(Number of people employed per km2),
Log(GVAdensity)= Log(Rand value of economic output per km2) Share_matric= Proportion of the population with matric
Level of significance *** =1%, **= 5%, *= 10%
APPENDICES
53
APPENDIX D
Built up area
N=27
Dependent variable
Popgrowth
Empt growth
GVA growth
-0.4850
-1.8719
-0.7234
(0.9425)
(1.5449)
(1.2148)
5.5485
18.2696
8.8643
(8.1158)
(13.3039)
(10.4612)
R
0.0105
0.0555
0.014
log(emptdensity)
0.4657
0.0323
0.4130
(0.6725)
(1.1390)
(0.8727)
-1.9518
1.9260
-0.3122
(4.8081)
(8.1441)
(6.2400)
0.0188
0
0.0089
log(popdensity)
Constant
2
Constant
R2
Notes: Bivariate regression takes the form y= b1x + u. Popgrowth, emptgrowth and GVAgrowth are calculated
using the equation growth= (ln (Xi10 /Xi96)/15)*100 Log(popdensity) = Log(Number of people per km2)
Log(emptdensity)= Log(Number of people employed per km2),
Level of significance *** =1%, **= 5%, *= 10%
54
APPENDICES
APPENDICES
55
0.8380
(0.7143)
(0.0069)
0.0371
gvagrowth
Popgrowth
0.5612
(1.9198)
0.7522
(0.0767)
0.1737*
0.2227
(7.4215)
-4.5024
(0.4232)
0.4531
0.1895
(2.1231)
1.3969
(0.3678)
0.3556
0.0523
(4.1594)
1.4874
(0.6002)
0.2822
0.0975
(1.5474)
-0.7840
(0.1273)
0.1823
0.0085
(7.4388)
-1.6133
(0.4877)
0.1974
0.0176
(1.7136)
2.3792
(0.4789)
-0.2793
0.0043
1.9996)
1.9586
(0.4065)
-0.1159
0.0654
(2.6312)
-0.8747
(0.2164)
0.2495
0.0032
(12.4642)
5.1607
(0.8171)
-0.2003
0.0587
(2.8029)
5.1133
(0.7832)
-0.8527
0.0139
(3.3252)
3.8064
(0.6760)
0.3493
emptgrowth
0.0122
(1.9842)
1.4639
(0.1632)
0.0790
0.0276
(9.0300)
9.0311
(0.5920)
-0.4345
0.0709
2.0426)
4.8313
(0.5708)
-0.6874
0.0120
(2.4413)
3.5687
(0.4963)
-0.2385
gvagrowth
0.1309
(0.3389)
0.0680
(0.4136)
0.1795
0.0499
(1.0185)
-1.1451
(0.0834)
0.2139
0.0502
(0.1431)
1.2598
(0.0893)
0.2295
0.0182
(0.2501)
1.1169
(0.0923)
0.1405
popgrowth
0.1769
(0.4491)
-0.4357
(0.0548)
0.2841***
0.0796
(1.3652)
-2.7084
(0.1118)
0.3676*
0.0775
(0.1921)
1.4310
(0.1198)
0.3881*
0.0645
(0.3325)
0.9019
(0.1227)
0.36*
emptgrowth
N=83
Commercial
0.1565
(0.3787)
0.5406
(0.0462)
0.2226***
0.0005
(1.1851)
1.9438
(0.0971)
0.0253
0.0025
(0.1664)
2.2029
(0.1038)
0.0580
0.0000
(0.2864)
2.2400
(0.1057)
0.0043
gvagrowth
0.0100
(0.3883)
0.0747
(0.0829)
0.0749
0.0013
(2.0562)
1.0542
(0.1604)
-0.0515
0.0010
(0.3054)
0.4699
(0.1909)
-0.0534
0.0310
(0.9088)
1.8388
(0.2201)
-0.3544
0.0005
(1.1934)
0.9627
(0.2550)
-0.0533
0.0273
(6.2076)
10.0735
(0.4842)
-0.7306
0.0133
(0.9285)
1.5679
(0.5806)
-0.6055
0.0181
(2.7986)
4.1038
(0.6778)
-0.828
0.1148
(0.3267)
0.7914
(6.9829)
0.2263***
0.0210
(1.8116)
-0.6153
(0.1413)
0.186
0.0161
(0.2697)
1.4956
(0.1686)
0.194
0.0022
(0.8206)
1.4228
(0.1988)
0.0836
popgrowth emptgrowth gvagrowth
N=127
Former bantustan
Notes: Bivariate regression takes the form y= b1x + u. Popgrowth, emptgrowth and GVAgrowth are calculated using the equation growth= (ln (Xi10 /Xi96)/15)*100 Log(popdensity) = Log(Number
of people per km2) Log(emptdensity)= Log(Number of people employed per km2), Log(GVAdensity)= Log(Rand value of economic output per km2) Share_matric= Proportion of the population
with matric
Level of significance *** =1%, **= 5%, *= 10%
R2
-0.8936
0.0103
(0.0457)
(0.0445)
Constant
0.2082*
0.0174
share_matric
0.4288
(6.2417)
(2.5487)
0.4020
-8.4831
-2.8765
(0.3559)
(0.1453)
0.4146
0.5081
0.6168
(1.7702)
(0.6476)
0.2383
-0.0636
(0.3066)
(0.1122)
-0.0087
0.5161
0.2629
0.5726
0.228
(3.5989)
(1.0937)
(0.5194)
(0.1578)
-1.9639
0.6204
0.3653*
-1.2269
emptgrowth
N=21
N=6
Popgrowth
Secondary
Metro
R2
Constant
log(gvadensity)
2
R
Constant
log(emptdensity)
2
R
Constant
log (popdensity)
APPENDIX E
56
APPENDICES
(2.4149)
0.0025
0.2469
(1.3404)
0.0088
(1.0565)
0.0275
0.0727
(0.1818
0.9809
(0.5954)
0.0039
log(emptdensity)
Constant
(1.2221)
0.2369
(0.5281)
0.2742
gvagrowth
0.3423
(0.4540)
0.9069
(0.0413)
0.1909***
0.0137
(2.2249)
1.1282
(0.1520)
0.1146
0.0114
(0.5356)
2.4670
(0.1635)
0.1124
0.0201
(0.9577)
3.6451
(0.1955)
-0.1793
0.1247
(0.2748)
0.0398
(0.0395)
0.1790***
0.1006
(1.0763)
-0.5934
(0.0885)
0.14634
0.0175
(0.1444)
1.0453
(0.0944)
0.1512
0.0034
(0.2653)
1.3993
(0.0874)
-0.0609
0.0415
(0.4984)
0.4267
(0.0716)
0.1790***
0.0277
(-2.1732)
-0.0220
(0.1527)
0.3090
0.0313
(0.2485)
1.2614
(0.1624)
0.3506
0.0076
(0.4587)
1.1323
(.15122)
0.1588
popgrowth emptgrowth
0.1386
(0.2297)
1.1659
(0.0330)
0.1589**
0.0050
(0.9131)
1.4033
(0.0751)
0.0637
0.0114
(0.1220)
2.0856
(0.0797)
0.1029
0.0022
(0.2236)
2.2866
(0.0737)
-0.0418
gvagrowth
0.1269
(0.9143)
-0.7311
(0.1036)
0.1894*
0.0371
(2.4108)
-1.3833
(0.1761)
0.1656
0.0061
(0.4054)
0.7593
(0.1695)
0.0639
0.0001
(0.6308)
0.8901
(0.1792)
-0.0065
popgrowth
0.1204
(2.4246)
-3.8662
(0.2748)
0.4875*
0.0037
(6.4793)
-1.6160
(0.4733)
0.1374
0.0021
(1.0733)
0.4218
(0.4490)
-0.0978
0.0001
(0.6308)
0.8901
(0.1792)
-0.0065
emptgrowth
N=25
Mining
0.0594
(1.6617)
-0.8766
(0.1883)
0.2268
0.0004
(4.3010)
1.4696
(0.3142)
-0.03156
0.0222
(0.7041)
1.4073
(0.2945)
-0.21282
0.0056
(1.1015)
1.3959
(0.3129)
-0.1128
gvagrowth
0.0065
(0.3906)
0.5239
(0.0701)
-0.2601
0.0034
(2.3021)
-0.2160
(0.1787)
0.0476
0.0000
(0.3521
0.4021
(0.2141)
-0.0046
0.2292
(1.5785)
4.3198
(0.3531)
-0.8824*
popgrowth
0.1381
(1.3468)
3.4780
(0.2417)
-0.4434
0.0463
(8.3361)
9.6773
(0.6471)
-0.6532
0.0686
(1.2578)
2.6291
(0.7651)
-0.9516
0.2379
(5.8100)
16.0906
(1.2996)
-3.3279
emptgrowth
N=23
0.1868
(0.4006)
0.9470
(0.0719)
0.1579**
0.1774
(2.3708)
-3.3036
(0.1840)
0.3917*
0.1378
(0.3706)
1.1451
(0.2254)
0.4130*
0.0079
(2.0300)
0.8989
(0.4541)
0.1860
Gvagrowth
Community Services
Notes: Bivariate regression takes the form y= b1x + u. Popgrowth, emptgrowth and GVAgrowth are calculated using the equation growth= (ln (Xi10 /Xi96)/15)*100 Log(popdensity) =
Log(Number of people per km2) Log(emptdensity)= Log(Number of people employed per km2), Log(GVAdensity)= Log(Rand value of economic output per km2) Share_matric= Proportion of the
population with matric
Level of significance *** =1%, **= 5%, *= 10%
R
-2.1371
-0.6784
Constant
2
(0.1112)
(0.0480)
0.3970***
-0.0041
0.1893***
0.0025
0.0016
share_matric
R
(5.5913)
(2.4787)
2
0.5648
Constant
(0.3820)
(0.1693)
0.1232
0.04320
2
1.0680
(0.4093)
log(gvadensity)
R
R
2.5357
2.2952
Constant
2
(0.4930)
(0.2157)
-0.1574
-0.2324
log (popdensity)
emptgrowth
N=146
N=43
popgrowth
Agriculture
Manufacturing
APPENDIX F
URBAN TRANSFORMATION RESEARCH PROJECT (UTRP)
The South African Research Chair in Development Planning and Modelling
School of Architecture and Planning
University of the Witwatersrand
Private Bag 3, Wits, 2050, Johannesburg, South Africa
http://www.wits.ac.za/academic/ebe/archplan/14488/sarchi.html
T +27 11 717 7642 • F +27 11 717 7624
Research Chair under the NRF South African Research Chairs Initiative (SARChI)
REHEMA MSULWA is currently pursuing a Masters degree in Applied Economics at the
University of Cape Town. She has interned with the Human Sciences Research Council
as a Junior Researcher and also with the Clinton Health Access Initiative. Her interests
include development economics and the design and implementation of development
projects in Africa. To date she has examined the spatial distribution of economic activity
in South Africa and healthcare accessibility in rural Tanzania.
PROFESSOR IVAN TUROK is Deputy Executive Director at the Human Sciences Research
Council, and Honorary Professor at the Universities of Glasgow and Cape Town. He is an
expert adviser to the United Nations, OECD, European Commission, SA Government,
UK Government and African Development Bank, and is a board member of the Regional
Studies Association. Ivan was the principal author of the 2011 State of South African
Cities Report. Other books include the State of English Cities (2006), Changing Cities:
Rethinking Urban Competitiveness, Cohesion and Governance (2005), Twin Track Cities
(2005), The Jobs Gap in Britain’s Cities (1999) and The Coherence of EU Regional Policy
(1997). He has written extensively on urban and regional development and policy in the
global north and is now committed to working in the south.
9 780992 177508
9 780992 177515
9 780992 177522
9 780992 177553
9 780992 177560
9 780992 177577
9 780992 17753
NRF9
Funded
780992
17758